def construct_axis_aligned_forest(self):
# (0) X[0] > 2.2
# / \
# (1) X[1] > -5 (2) [0.7 0.1 0.2]
# / \
# (3) [0.3 0.3 0.4] (4) [0.3 0.6 0.1]
#
# (0) X[0] > 5.0
# / \
# (1) X[0] > 2.5 (2) [0.8 0.1 0.1]
# / \
# (3) [0.2 0.2 0.6] (4) [0.2 0.7 0.1]
path_1 = buffers.as_matrix_buffer(np.array([[1,2],[3,4],[-1,-1],[-1,-1],[-1,-1]], dtype=np.int32))
int_params_1 = buffers.as_matrix_buffer(np.array([[1,0],[1,1],[1,0],[1,0],[1,0]], dtype=np.int32))
float_params_1 = buffers.as_matrix_buffer(np.array([[2.2],[-5],[0],[0],[0]], dtype=np.float32))
ys_1 = buffers.as_matrix_buffer(np.array([[0,0,0],[0,0,0],[0.7,0.1,0.2],[0.3,0.3,0.4],[0.3,0.6,0.1]], dtype=np.float32))
depth_1 = buffers.as_vector_buffer(np.array([0, 1, 1, 2, 2], dtype=np.int32))
counts_1 = buffers.as_vector_buffer(np.array([5, 5, 5, 5, 5], dtype=np.float32))
tree_1 = forest_data.Tree(path_1, int_params_1, float_params_1, depth_1, counts_1, ys_1)
path_2 = buffers.as_matrix_buffer(np.array([[1,2],[3,4],[-1,-1],[-1,-1],[-1,-1]], dtype=np.int32))
int_params_2 = buffers.as_matrix_buffer(np.array([[1,0],[1,0],[1,0],[1,0],[1,0]], dtype=np.int32))
float_params_2 = buffers.as_matrix_buffer(np.array([[5.0],[2.5],[0],[0],[0]], dtype=np.float32))
ys_2 = buffers.as_matrix_buffer(np.array([[0,0,0],[0,0,0],[0.8,0.1,0.1],[0.2,0.2,0.6],[0.2,0.7,0.1]], dtype=np.float32))
depth_2 = buffers.as_vector_buffer(np.array([0, 1, 1, 2, 2], dtype=np.int32))
counts_2 = buffers.as_vector_buffer(np.array([5, 5, 5, 5, 5], dtype=np.float32))
tree_2 = forest_data.Tree(path_2, int_params_2, float_params_2, depth_2, counts_2, ys_2)
forest = forest_data.Forest([tree_1, tree_2])
return forest
def construct_axis_aligned_forest(self):
# (0) X[0] > 2.2
# / \
# (1) X[1] > -5 (2) [0.7 0.1 0.2]
# / \
# (3) [0.3 0.3 0.4] (4) [0.3 0.6 0.1]
#
# (0) X[0] > 5.0
# / \
# (1) X[0] > 2.5 (2) [0.8 0.1 0.1]
# / \
# (3) [0.2 0.2 0.6] (4) [0.2 0.7 0.1]
path_1 = buffers.as_matrix_buffer(
np.array([[1, 2], [3, 4], [-1, -1], [-1, -1], [-1, -1]],
dtype=np.int32))
int_params_1 = buffers.as_matrix_buffer(
np.array([[1, 0], [1, 1], [1, 0], [1, 0], [1, 0]], dtype=np.int32))
float_params_1 = buffers.as_matrix_buffer(
np.array([[2.2], [-5], [0], [0], [0]], dtype=np.float32))
ys_1 = buffers.as_matrix_buffer(
np.array([[0, 0, 0], [0, 0, 0], [0.7, 0.1, 0.2], [0.3, 0.3, 0.4],
[0.3, 0.6, 0.1]],
dtype=np.float32))
depth_1 = buffers.as_vector_buffer(
np.array([0, 1, 1, 2, 2], dtype=np.int32))
counts_1 = buffers.as_vector_buffer(
np.array([5, 5, 5, 5, 5], dtype=np.float32))
tree_1 = forest_data.Tree(path_1, int_params_1, float_params_1,
depth_1, counts_1, ys_1)
path_2 = buffers.as_matrix_buffer(
np.array([[1, 2], [3, 4], [-1, -1], [-1, -1], [-1, -1]],
dtype=np.int32))
int_params_2 = buffers.as_matrix_buffer(
np.array([[1, 0], [1, 0], [1, 0], [1, 0], [1, 0]], dtype=np.int32))
float_params_2 = buffers.as_matrix_buffer(
np.array([[5.0], [2.5], [0], [0], [0]], dtype=np.float32))
ys_2 = buffers.as_matrix_buffer(
np.array([[0, 0, 0], [0, 0, 0], [0.8, 0.1, 0.1], [0.2, 0.2, 0.6],
[0.2, 0.7, 0.1]],
dtype=np.float32))
depth_2 = buffers.as_vector_buffer(
np.array([0, 1, 1, 2, 2], dtype=np.int32))
counts_2 = buffers.as_vector_buffer(
np.array([5, 5, 5, 5, 5], dtype=np.float32))
tree_2 = forest_data.Tree(path_2, int_params_2, float_params_2,
depth_2, counts_2, ys_2)
tree_2.GetExtraInfo().AddBuffer(
"first", np.array([3, 21, 1, 22, 1, 5], dtype=np.float32))
forest = forest_data.Forest([tree_1, tree_2])
return forest
def create_scaled_depth_delta_learner_32f(**kwargs):
ux = float( kwargs.get('ux') )
uy = float( kwargs.get('uy') )
vx = float( kwargs.get('vx') )
vy = float( kwargs.get('vy') )
number_of_trees = int( kwargs.get('number_of_trees', 10) )
number_of_features = int( kwargs.get('number_of_features', 1) )
feature_ordering = int( kwargs.get('feature_ordering', pipeline.FEATURES_BY_DATAPOINTS) )
number_of_jobs = int( kwargs.get('number_of_jobs', 1) )
number_of_classes = int( kwargs['classes'].GetMax() + 1 )
try_split_criteria = create_try_split_criteria(**kwargs)
if 'bootstrap' in kwargs and kwargs.get('bootstrap'):
sample_data_step = pipeline.BootstrapSamplesStep_i32f32i32(buffers.PIXEL_INDICES)
else:
sample_data_step = pipeline.AllSamplesStep_i32f32i32(buffers.PIXEL_INDICES)
number_of_features_buffer = buffers.as_vector_buffer(np.array([number_of_features], dtype=np.int32))
set_number_features_step = pipeline.SetInt32VectorBufferStep(number_of_features_buffer, pipeline.WHEN_NEW)
tree_steps_pipeline = pipeline.Pipeline([sample_data_step, set_number_features_step])
feature_params_step = image_features.PixelPairGaussianOffsetsStep_f32i32(set_number_features_step.OutputBufferId, ux, uy, vx, vy )
depth_delta_feature = image_features.ScaledDepthDeltaFeature_f32i32(feature_params_step.FloatParamsBufferId,
feature_params_step.IntParamsBufferId,
sample_data_step.IndicesBufferId,
buffers.PIXEL_INDICES,
buffers.DEPTH_IMAGES,
buffers.OFFSET_SCALES)
depth_delta_feature_extractor_step = image_features.ScaledDepthDeltaFeatureExtractorStep_f32i32(depth_delta_feature, feature_ordering)
slice_classes_step = pipeline.SliceInt32VectorBufferStep_i32(buffers.CLASS_LABELS, sample_data_step.IndicesBufferId)
slice_weights_step = pipeline.SliceFloat32VectorBufferStep_i32(sample_data_step.WeightsBufferId, sample_data_step.IndicesBufferId)
class_infogain_walker = classification.ClassInfoGainWalker_f32i32(slice_weights_step.SlicedBufferId,
slice_classes_step.SlicedBufferId,
number_of_classes)
best_splitpint_step = classification.ClassInfoGainBestSplitpointsWalkingSortedStep_f32i32(class_infogain_walker,
depth_delta_feature_extractor_step.FeatureValuesBufferId,
feature_ordering)
node_steps_pipeline = pipeline.Pipeline([feature_params_step, depth_delta_feature_extractor_step,
slice_classes_step, slice_weights_step, best_splitpint_step])
split_buffers = splitpoints.SplitSelectorBuffers(best_splitpint_step.ImpurityBufferId,
best_splitpint_step.SplitpointBufferId,
best_splitpint_step.SplitpointCountsBufferId,
best_splitpint_step.ChildCountsBufferId,
best_splitpint_step.LeftYsBufferId,
best_splitpint_step.RightYsBufferId,
feature_params_step.FloatParamsBufferId,
feature_params_step.IntParamsBufferId,
depth_delta_feature_extractor_step.FeatureValuesBufferId,
feature_ordering,
sample_data_step.IndicesBufferId)
should_split_criteria = create_should_split_criteria(**kwargs)
finalizer = classification.ClassEstimatorFinalizer_f32()
split_selector = splitpoints.SplitSelector_f32i32([split_buffers], should_split_criteria, finalizer )
tree_learner = learn.DepthFirstTreeLearner_f32i32(try_split_criteria, tree_steps_pipeline, node_steps_pipeline, split_selector)
forest_learner = learn.ParallelForestLearner(tree_learner, number_of_trees, 5, 5, number_of_classes, number_of_jobs)
return forest_learner
def test_pickle(self):
# test tensor
for dtype in [np.float32, np.float64, np.int32, np.int64]:
array = np.array(
[[[3, 21, 1], [22, 1, 5]], [[2, 2, 2], [7, 7, 7]]],
dtype=dtype)
b1 = buffers.as_tensor_buffer(array)
pickle.dump(b1, open('tmp.pkl', 'wb'))
b2 = pickle.load(open('tmp.pkl', 'rb'))
array2 = buffers.as_numpy_array(b2)
self.assertTrue((array == array2).all())
# test matrix
for dtype in [np.float32, np.float64, np.int32, np.int64]:
array = np.array([[3, 21, 1], [22, 1, 5], [2, 2, 2], [7, 7, 7]],
dtype=dtype)
b1 = buffers.as_matrix_buffer(array)
pickle.dump(b1, open('tmp.pkl', 'wb'))
b2 = pickle.load(open('tmp.pkl', 'rb'))
array2 = buffers.as_numpy_array(b2)
self.assertTrue((array == array2).all())
# test vector
for dtype in [np.float32, np.float64, np.int32, np.int64]:
array = np.array([3, 21, 1], dtype=dtype)
b1 = buffers.as_vector_buffer(array)
pickle.dump(b1, open('tmp.pkl', 'wb'))
b2 = pickle.load(open('tmp.pkl', 'rb'))
array2 = buffers.as_numpy_array(b2)
self.assertTrue((array == array2).all())
def test_pickle(self):
# test tensor
for dtype in [np.float32, np.float64, np.int32, np.int64]:
array = np.array([[[3, 21, 1], [22, 1, 5]], [[2, 2, 2], [7, 7, 7]]], dtype=dtype)
b1 = buffers.as_tensor_buffer(array)
pickle.dump(b1, open("tmp.pkl", "wb"))
b2 = pickle.load(open("tmp.pkl", "rb"))
array2 = buffers.as_numpy_array(b2)
self.assertTrue((array == array2).all())
# test matrix
for dtype in [np.float32, np.float64, np.int32, np.int64]:
array = np.array([[3, 21, 1], [22, 1, 5], [2, 2, 2], [7, 7, 7]], dtype=dtype)
b1 = buffers.as_matrix_buffer(array)
pickle.dump(b1, open("tmp.pkl", "wb"))
b2 = pickle.load(open("tmp.pkl", "rb"))
array2 = buffers.as_numpy_array(b2)
self.assertTrue((array == array2).all())
# test vector
for dtype in [np.float32, np.float64, np.int32, np.int64]:
array = np.array([3, 21, 1], dtype=dtype)
b1 = buffers.as_vector_buffer(array)
pickle.dump(b1, open("tmp.pkl", "wb"))
b2 = pickle.load(open("tmp.pkl", "rb"))
array2 = buffers.as_numpy_array(b2)
self.assertTrue((array == array2).all())
def create_biau2008_regression_axis_aligned_matrix_learner_32f(**kwargs):
number_of_trees = int( kwargs.get('number_of_trees', 10) )
number_of_leaves = int( kwargs.get('number_of_leaves', kwargs['y'].shape[0] / 5 + 1) )
number_of_split_retries = int( kwargs.get('number_of_split_retries', 10) )
number_of_features = 1
feature_ordering = int( kwargs.get('feature_ordering', pipeline.FEATURES_BY_DATAPOINTS) )
number_of_jobs = int( kwargs.get('number_of_jobs', 1) )
dimension_of_y = int( kwargs['y'].shape[1] )
try_split_criteria = try_split.MinNodeSizeCriteria(2)
sample_data_step = pipeline.AllSamplesStep_f32f32i32(buffers.X_FLOAT_DATA)
number_of_features_buffer = buffers.as_vector_buffer(np.array([number_of_features], dtype=np.int32))
set_number_features_step = pipeline.SetInt32VectorBufferStep(number_of_features_buffer, pipeline.WHEN_NEW)
tree_steps_pipeline = pipeline.Pipeline([sample_data_step, set_number_features_step])
feature_params_step = matrix_features.AxisAlignedParamsStep_f32i32(set_number_features_step.OutputBufferId, buffers.X_FLOAT_DATA)
matrix_feature = matrix_features.LinearFloat32MatrixFeature_f32i32(feature_params_step.FloatParamsBufferId,
feature_params_step.IntParamsBufferId,
sample_data_step.IndicesBufferId,
buffers.X_FLOAT_DATA)
matrix_feature_extractor_step = matrix_features.LinearFloat32MatrixFeatureExtractorStep_f32i32(matrix_feature, feature_ordering)
slice_ys_step = pipeline.SliceFloat32MatrixBufferStep_i32(buffers.YS, sample_data_step.IndicesBufferId)
slice_weights_step = pipeline.SliceFloat32VectorBufferStep_i32(sample_data_step.WeightsBufferId, sample_data_step.IndicesBufferId)
impurity_walker = regression.SumOfVarianceWalker_f32i32(slice_weights_step.SlicedBufferId,
slice_ys_step.SlicedBufferId,
dimension_of_y)
best_splitpint_step = regression.SumOfVarianceRandomGapSplitpointsStep_f32i32(impurity_walker,
matrix_feature_extractor_step.FeatureValuesBufferId,
feature_ordering)
node_steps_pipeline = pipeline.Pipeline([feature_params_step, matrix_feature_extractor_step,
slice_ys_step, slice_weights_step, best_splitpint_step])
split_buffers = splitpoints.SplitSelectorBuffers(best_splitpint_step.ImpurityBufferId,
best_splitpint_step.SplitpointBufferId,
best_splitpint_step.SplitpointCountsBufferId,
best_splitpint_step.ChildCountsBufferId,
best_splitpint_step.LeftYsBufferId,
best_splitpint_step.RightYsBufferId,
feature_params_step.FloatParamsBufferId,
feature_params_step.IntParamsBufferId,
matrix_feature_extractor_step.FeatureValuesBufferId,
feature_ordering,
matrix_feature_extractor_step)
should_split_criteria = min_child_size_criteria = should_split.MinChildSizeCriteria(1)
finalizer = regression.MeanVarianceEstimatorFinalizer_f32()
split_indices = splitpoints.SplitIndices_f32i32(sample_data_step.IndicesBufferId)
split_selector = splitpoints.SplitSelector_f32i32([split_buffers], should_split_criteria, finalizer, split_indices )
tree_learner = learn.Biau2008TreeLearner_f32i32(try_split_criteria, tree_steps_pipeline, node_steps_pipeline, split_selector, number_of_leaves, number_of_split_retries)
forest_learner = learn.ParallelForestLearner(tree_learner, number_of_trees, dimension_of_y, number_of_jobs)
return forest_learner
def load_and_sample(pose_path, list_of_poses, number_of_pixels_per_image):
depths, labels = load_data(pose_path, list_of_poses)
depths_buffer = buffers.as_tensor_buffer(depths)
del depths
pixel_indices, pixel_labels = sample_pixels_from_images(labels, number_of_pixels_per_image)
del labels
pixel_indices_buffer = buffers.as_matrix_buffer(pixel_indices)
pixel_labels_buffer = buffers.as_vector_buffer(pixel_labels)
del pixel_indices
del pixel_labels
return depths_buffer, pixel_indices_buffer, pixel_labels_buffer
def load_training_data(numpy_filename):
f = open(numpy_filename, 'rb')
depths = np.load(f)
labels = np.load(f)
pixel_indices = np.load(f)
pixel_labels = np.load(f)
depths_buffer = buffers.as_tensor_buffer(depths)
del depths
del labels
pixel_indices_buffer = buffers.as_matrix_buffer(pixel_indices)
del pixel_indices
pixel_labels_buffer = buffers.as_vector_buffer(pixel_labels)
del pixel_labels
return depths_buffer, pixel_indices_buffer, pixel_labels_buffer
def create_axis_aligned_matrix_two_stream_learner_32f(**kwargs):
number_of_trees = int(kwargs.get('number_of_trees', 10))
number_of_features = int(
kwargs.get('number_of_features', np.sqrt(kwargs['x'].shape[1])))
feature_ordering = int(
kwargs.get('feature_ordering', pipeline.FEATURES_BY_DATAPOINTS))
number_of_splitpoints = int(kwargs.get('number_of_splitpoints', 1))
number_of_jobs = int(kwargs.get('number_of_jobs', 1))
number_of_classes = int(np.max(kwargs['classes']) + 1)
probability_of_impurity_stream = float(
kwargs.get('probability_of_impurity_stream', 0.5))
try_split_criteria = create_try_split_criteria(**kwargs)
if 'bootstrap' in kwargs and kwargs.get('bootstrap'):
sample_data_step = pipeline.BootstrapSamplesStep_f32f32i32(
buffers.X_FLOAT_DATA)
else:
sample_data_step = pipeline.AllSamplesStep_f32f32i32(
buffers.X_FLOAT_DATA)
number_of_features_buffer = buffers.as_vector_buffer(
np.array([number_of_features], dtype=np.int32))
set_number_features_step = pipeline.SetInt32VectorBufferStep(
number_of_features_buffer, pipeline.WHEN_NEW)
assign_stream_step = splitpoints.AssignStreamStep_f32i32(
sample_data_step.WeightsBufferId, probability_of_impurity_stream)
tree_steps_pipeline = pipeline.Pipeline(
[sample_data_step, set_number_features_step, assign_stream_step])
feature_params_step = matrix_features.AxisAlignedParamsStep_f32i32(
set_number_features_step.OutputBufferId, buffers.X_FLOAT_DATA)
matrix_feature = matrix_features.LinearFloat32MatrixFeature_f32i32(
feature_params_step.FloatParamsBufferId,
feature_params_step.IntParamsBufferId,
sample_data_step.IndicesBufferId, buffers.X_FLOAT_DATA)
matrix_feature_extractor_step = matrix_features.LinearFloat32MatrixFeatureExtractorStep_f32i32(
matrix_feature, feature_ordering)
slice_classes_step = pipeline.SliceInt32VectorBufferStep_i32(
buffers.CLASS_LABELS, sample_data_step.IndicesBufferId)
slice_weights_step = pipeline.SliceFloat32VectorBufferStep_i32(
sample_data_step.WeightsBufferId, sample_data_step.IndicesBufferId)
slice_assign_stream_step = pipeline.SliceInt32VectorBufferStep_i32(
assign_stream_step.StreamTypeBufferId,
sample_data_step.IndicesBufferId)
random_splitpoint_selection_step = splitpoints.RandomSplitpointsStep_f32i32(
matrix_feature_extractor_step.FeatureValuesBufferId,
number_of_splitpoints, feature_ordering,
slice_assign_stream_step.SlicedBufferId)
class_stats_updater = classification.ClassStatsUpdater_f32i32(
slice_weights_step.SlicedBufferId, slice_classes_step.SlicedBufferId,
number_of_classes)
two_stream_split_stats_step = classification.ClassStatsUpdaterTwoStreamStep_f32i32(
random_splitpoint_selection_step.SplitpointsBufferId,
random_splitpoint_selection_step.SplitpointsCountsBufferId,
slice_assign_stream_step.SlicedBufferId,
matrix_feature_extractor_step.FeatureValuesBufferId, feature_ordering,
class_stats_updater)
impurity_step = classification.ClassInfoGainSplitpointsImpurity_f32i32(
random_splitpoint_selection_step.SplitpointsCountsBufferId,
two_stream_split_stats_step.ChildCountsImpurityBufferId,
two_stream_split_stats_step.LeftImpurityStatsBufferId,
two_stream_split_stats_step.RightImpurityStatsBufferId)
node_steps_pipeline = pipeline.Pipeline([
feature_params_step, matrix_feature_extractor_step, slice_classes_step,
slice_weights_step, slice_assign_stream_step,
random_splitpoint_selection_step, two_stream_split_stats_step,
impurity_step
])
split_buffers = splitpoints.SplitSelectorBuffers(
impurity_step.ImpurityBufferId,
random_splitpoint_selection_step.SplitpointsBufferId,
random_splitpoint_selection_step.SplitpointsCountsBufferId,
two_stream_split_stats_step.ChildCountsEstimatorBufferId,
two_stream_split_stats_step.LeftEstimatorStatsBufferId,
two_stream_split_stats_step.RightEstimatorStatsBufferId,
feature_params_step.FloatParamsBufferId,
feature_params_step.IntParamsBufferId,
matrix_feature_extractor_step.FeatureValuesBufferId, feature_ordering,
matrix_feature_extractor_step)
should_split_criteria = create_should_split_criteria(**kwargs)
finalizer = classification.ClassEstimatorFinalizer_f32()
split_indices = splitpoints.SplitIndices_f32i32(
sample_data_step.IndicesBufferId)
split_selector = splitpoints.SplitSelector_f32i32([split_buffers],
should_split_criteria,
finalizer, split_indices)
if 'tree_order' in kwargs and kwargs.get('tree_order') == 'breadth_first':
tree_learner = learn.BreadthFirstTreeLearner_f32i32(
try_split_criteria, tree_steps_pipeline, node_steps_pipeline,
split_selector)
else:
tree_learner = learn.DepthFirstTreeLearner_f32i32(
try_split_criteria, tree_steps_pipeline, node_steps_pipeline,
split_selector)
forest_learner = learn.ParallelForestLearner(tree_learner, number_of_trees,
number_of_classes,
number_of_jobs)
return forest_learner
def create_class_pair_difference_matrix_walking_learner_32f(**kwargs):
number_of_trees = int(kwargs.get('number_of_trees', 10))
number_of_features = int(
kwargs.get('number_of_features', np.sqrt(kwargs['x'].shape[1])))
feature_ordering = int(
kwargs.get('feature_ordering', pipeline.FEATURES_BY_DATAPOINTS))
number_of_jobs = int(kwargs.get('number_of_jobs', 1))
number_of_classes = int(np.max(kwargs['classes']) + 1)
try_split_criteria = create_try_split_criteria(**kwargs)
if 'bootstrap' in kwargs and kwargs.get('bootstrap'):
sample_data_step = pipeline.BootstrapSamplesStep_f32f32i32(
buffers.X_FLOAT_DATA)
else:
sample_data_step = pipeline.AllSamplesStep_f32f32i32(
buffers.X_FLOAT_DATA)
number_of_features_buffer = buffers.as_vector_buffer(
np.array([number_of_features], dtype=np.int32))
set_number_features_step = pipeline.SetInt32VectorBufferStep(
number_of_features_buffer, pipeline.WHEN_NEW)
tree_steps_pipeline = pipeline.Pipeline(
[sample_data_step, set_number_features_step])
feature_params_step = matrix_features.ClassPairDifferenceParamsStep_f32i32(
set_number_features_step.OutputBufferId, buffers.X_FLOAT_DATA,
buffers.CLASS_LABELS, sample_data_step.IndicesBufferId)
matrix_feature = matrix_features.LinearFloat32MatrixFeature_f32i32(
feature_params_step.FloatParamsBufferId,
feature_params_step.IntParamsBufferId,
sample_data_step.IndicesBufferId, buffers.X_FLOAT_DATA)
matrix_feature_extractor_step = matrix_features.LinearFloat32MatrixFeatureExtractorStep_f32i32(
matrix_feature, feature_ordering)
slice_classes_step = pipeline.SliceInt32VectorBufferStep_i32(
buffers.CLASS_LABELS, sample_data_step.IndicesBufferId)
slice_weights_step = pipeline.SliceFloat32VectorBufferStep_i32(
sample_data_step.WeightsBufferId, sample_data_step.IndicesBufferId)
class_infogain_walker = classification.ClassInfoGainWalker_f32i32(
slice_weights_step.SlicedBufferId, slice_classes_step.SlicedBufferId,
number_of_classes)
best_splitpint_step = classification.ClassInfoGainBestSplitpointsWalkingSortedStep_f32i32(
class_infogain_walker,
matrix_feature_extractor_step.FeatureValuesBufferId, feature_ordering)
node_steps_pipeline = pipeline.Pipeline([
feature_params_step, matrix_feature_extractor_step, slice_classes_step,
slice_weights_step, best_splitpint_step
])
split_buffers = splitpoints.SplitSelectorBuffers(
best_splitpint_step.ImpurityBufferId,
best_splitpint_step.SplitpointBufferId,
best_splitpint_step.SplitpointCountsBufferId,
best_splitpint_step.ChildCountsBufferId,
best_splitpint_step.LeftYsBufferId,
best_splitpint_step.RightYsBufferId,
feature_params_step.FloatParamsBufferId,
feature_params_step.IntParamsBufferId,
matrix_feature_extractor_step.FeatureValuesBufferId, feature_ordering,
matrix_feature_extractor_step)
should_split_criteria = create_should_split_criteria(**kwargs)
finalizer = classification.ClassEstimatorFinalizer_f32()
split_indices = splitpoints.SplitIndices_f32i32(
sample_data_step.IndicesBufferId)
split_selector = splitpoints.SplitSelector_f32i32([split_buffers],
should_split_criteria,
finalizer, split_indices)
if 'tree_order' in kwargs and kwargs.get('tree_order') == 'breadth_first':
tree_learner = learn.BreadthFirstTreeLearner_f32i32(
try_split_criteria, tree_steps_pipeline, node_steps_pipeline,
split_selector)
else:
tree_learner = learn.DepthFirstTreeLearner_f32i32(
try_split_criteria, tree_steps_pipeline, node_steps_pipeline,
split_selector)
forest_learner = learn.ParallelForestLearner(tree_learner, number_of_trees,
number_of_classes,
number_of_jobs)
return forest_learner
def create_biau2012_regression_axis_aligned_matrix_learner_32f(**kwargs):
number_of_trees = int(kwargs.get('number_of_trees', 10))
number_of_leaves = int(
kwargs.get('number_of_leaves', kwargs['y'].shape[0] / 5 + 1))
number_of_features = int(
kwargs.get('number_of_features', (kwargs['x'].shape[1]) / 3 + 0.5))
# number_of_features = int( kwargs.get('number_of_features', np.sqrt(kwargs['x'].shape[1])))
feature_ordering = int(
kwargs.get('feature_ordering', pipeline.FEATURES_BY_DATAPOINTS))
number_of_jobs = int(kwargs.get('number_of_jobs', 1))
dimension_of_y = int(kwargs['y'].shape[1])
probability_of_impurity_stream = float(
kwargs.get('probability_of_impurity_stream', 0.5))
try_split_criteria = create_try_split_criteria(**kwargs)
sample_data_step = pipeline.AllSamplesStep_f32f32i32(buffers.X_FLOAT_DATA)
number_of_features_buffer = buffers.as_vector_buffer(
np.array([number_of_features], dtype=np.int32))
set_number_features_step = pipeline.SetInt32VectorBufferStep(
number_of_features_buffer, pipeline.WHEN_NEW)
assert (
np.max(np.abs(kwargs['x'])) <= 1.00
) # double check that the data has been scaled into a -1,1 hypercube
feature_range_buffer = buffers.as_vector_buffer(
np.array([-1, 1], dtype=np.float32))
set_feature_range_buffer_step = pipeline.SetFloat32VectorBufferStep(
feature_range_buffer, pipeline.WHEN_NEW)
assign_stream_step = splitpoints.AssignStreamStep_f32i32(
sample_data_step.WeightsBufferId, probability_of_impurity_stream,
False)
forest_steps_pipeline = pipeline.Pipeline([
sample_data_step, set_number_features_step,
set_feature_range_buffer_step, assign_stream_step
])
tree_steps_pipeline = pipeline.Pipeline([])
feature_params_step = matrix_features.AxisAlignedParamsStep_f32i32(
set_number_features_step.OutputBufferId, buffers.X_FLOAT_DATA)
matrix_feature = matrix_features.LinearFloat32MatrixFeature_f32i32(
feature_params_step.FloatParamsBufferId,
feature_params_step.IntParamsBufferId,
sample_data_step.IndicesBufferId, buffers.X_FLOAT_DATA)
matrix_feature_extractor_step = matrix_features.LinearFloat32MatrixFeatureExtractorStep_f32i32(
matrix_feature, feature_ordering)
slice_ys_step = pipeline.SliceFloat32MatrixBufferStep_i32(
buffers.YS, sample_data_step.IndicesBufferId)
slice_weights_step = pipeline.SliceFloat32VectorBufferStep_i32(
sample_data_step.WeightsBufferId, sample_data_step.IndicesBufferId)
slice_assign_stream_step = pipeline.SliceInt32VectorBufferStep_i32(
assign_stream_step.StreamTypeBufferId,
sample_data_step.IndicesBufferId)
quantized_feature_equal = pipeline.FeatureEqualQuantized_f32i32(1.0)
midpoint_step = splitpoints.RangeMidpointStep_f32i32(
feature_params_step.FloatParamsBufferId,
feature_params_step.IntParamsBufferId,
set_feature_range_buffer_step.OutputBufferId, quantized_feature_equal)
mean_variance_stats_updater = regression.MeanVarianceStatsUpdater_f32i32(
slice_weights_step.SlicedBufferId, slice_ys_step.SlicedBufferId,
dimension_of_y)
two_stream_split_stats_step = regression.SumOfVarianceTwoStreamStep_f32i32(
midpoint_step.SplitpointsBufferId,
midpoint_step.SplitpointsCountsBufferId,
slice_assign_stream_step.SlicedBufferId,
matrix_feature_extractor_step.FeatureValuesBufferId, feature_ordering,
mean_variance_stats_updater)
impurity_step = regression.SumOfVarianceSplitpointsImpurity_f32i32(
midpoint_step.SplitpointsCountsBufferId,
two_stream_split_stats_step.ChildCountsImpurityBufferId,
two_stream_split_stats_step.LeftImpurityStatsBufferId,
two_stream_split_stats_step.RightImpurityStatsBufferId)
node_steps_pipeline = pipeline.Pipeline([
feature_params_step, matrix_feature_extractor_step, slice_ys_step,
slice_weights_step, slice_assign_stream_step, midpoint_step,
two_stream_split_stats_step, impurity_step
])
split_buffers = splitpoints.SplitSelectorBuffers(
impurity_step.ImpurityBufferId, midpoint_step.SplitpointsBufferId,
midpoint_step.SplitpointsCountsBufferId,
two_stream_split_stats_step.ChildCountsEstimatorBufferId,
two_stream_split_stats_step.LeftEstimatorStatsBufferId,
two_stream_split_stats_step.RightEstimatorStatsBufferId,
feature_params_step.FloatParamsBufferId,
feature_params_step.IntParamsBufferId,
matrix_feature_extractor_step.FeatureValuesBufferId, feature_ordering,
matrix_feature_extractor_step)
should_split_criteria = no_split_criteria(**kwargs)
finalizer = regression.MeanVarianceEstimatorFinalizer_f32()
split_indices = splitpoints.SplitIndices_f32i32(
sample_data_step.IndicesBufferId)
split_midpoint_ranges = splitpoints.SplitBuffersFeatureRange_f32i32(
midpoint_step.PastFloatParamsBufferId,
midpoint_step.PastIntParamsBufferId, midpoint_step.PastRangesBufferId,
set_feature_range_buffer_step.OutputBufferId, quantized_feature_equal)
split_steps = splitpoints.SplitBuffersList(
[split_indices, split_midpoint_ranges])
split_selector = splitpoints.SplitSelector_f32i32([split_buffers],
should_split_criteria,
finalizer, split_steps)
tree_learner = learn.BreadthFirstTreeLearner_f32i32(
try_split_criteria, tree_steps_pipeline, node_steps_pipeline,
split_selector, number_of_leaves)
forest_learner = learn.ParallelForestLearner(tree_learner,
forest_steps_pipeline,
number_of_trees,
dimension_of_y,
number_of_jobs)
return forest_learner
def create_axis_aligned_matrix_two_stream_learner_32f(**kwargs):
number_of_trees = int(kwargs.get("number_of_trees", 10))
number_of_features = int(kwargs.get("number_of_features", np.sqrt(kwargs["x"].shape[1])))
feature_ordering = int(kwargs.get("feature_ordering", pipeline.FEATURES_BY_DATAPOINTS))
number_of_splitpoints = int(kwargs.get("number_of_splitpoints", 1))
number_of_jobs = int(kwargs.get("number_of_jobs", 1))
number_of_classes = int(np.max(kwargs["classes"]) + 1)
probability_of_impurity_stream = float(kwargs.get("probability_of_impurity_stream", 0.5))
try_split_criteria = create_try_split_criteria(**kwargs)
if "bootstrap" in kwargs and kwargs.get("bootstrap"):
sample_data_step = pipeline.BootstrapSamplesStep_f32f32i32(buffers.X_FLOAT_DATA)
else:
sample_data_step = pipeline.AllSamplesStep_f32f32i32(buffers.X_FLOAT_DATA)
number_of_features_buffer = buffers.as_vector_buffer(np.array([number_of_features], dtype=np.int32))
set_number_features_step = pipeline.SetInt32VectorBufferStep(number_of_features_buffer, pipeline.WHEN_NEW)
assign_stream_step = splitpoints.AssignStreamStep_f32i32(
sample_data_step.WeightsBufferId, probability_of_impurity_stream
)
tree_steps_pipeline = pipeline.Pipeline([sample_data_step, set_number_features_step, assign_stream_step])
feature_params_step = matrix_features.AxisAlignedParamsStep_f32i32(
set_number_features_step.OutputBufferId, buffers.X_FLOAT_DATA
)
matrix_feature = matrix_features.LinearFloat32MatrixFeature_f32i32(
feature_params_step.FloatParamsBufferId,
feature_params_step.IntParamsBufferId,
sample_data_step.IndicesBufferId,
buffers.X_FLOAT_DATA,
)
matrix_feature_extractor_step = matrix_features.LinearFloat32MatrixFeatureExtractorStep_f32i32(
matrix_feature, feature_ordering
)
slice_classes_step = pipeline.SliceInt32VectorBufferStep_i32(buffers.CLASS_LABELS, sample_data_step.IndicesBufferId)
slice_weights_step = pipeline.SliceFloat32VectorBufferStep_i32(
sample_data_step.WeightsBufferId, sample_data_step.IndicesBufferId
)
slice_assign_stream_step = pipeline.SliceInt32VectorBufferStep_i32(
assign_stream_step.StreamTypeBufferId, sample_data_step.IndicesBufferId
)
random_splitpoint_selection_step = splitpoints.RandomSplitpointsStep_f32i32(
matrix_feature_extractor_step.FeatureValuesBufferId,
number_of_splitpoints,
feature_ordering,
slice_assign_stream_step.SlicedBufferId,
)
class_stats_updater = classification.ClassStatsUpdater_f32i32(
slice_weights_step.SlicedBufferId, slice_classes_step.SlicedBufferId, number_of_classes
)
two_stream_split_stats_step = classification.ClassStatsUpdaterTwoStreamStep_f32i32(
random_splitpoint_selection_step.SplitpointsBufferId,
random_splitpoint_selection_step.SplitpointsCountsBufferId,
slice_assign_stream_step.SlicedBufferId,
matrix_feature_extractor_step.FeatureValuesBufferId,
feature_ordering,
class_stats_updater,
)
impurity_step = classification.ClassInfoGainSplitpointsImpurity_f32i32(
random_splitpoint_selection_step.SplitpointsCountsBufferId,
two_stream_split_stats_step.ChildCountsImpurityBufferId,
two_stream_split_stats_step.LeftImpurityStatsBufferId,
two_stream_split_stats_step.RightImpurityStatsBufferId,
)
node_steps_pipeline = pipeline.Pipeline(
[
feature_params_step,
matrix_feature_extractor_step,
slice_classes_step,
slice_weights_step,
slice_assign_stream_step,
random_splitpoint_selection_step,
two_stream_split_stats_step,
impurity_step,
]
)
split_buffers = splitpoints.SplitSelectorBuffers(
impurity_step.ImpurityBufferId,
random_splitpoint_selection_step.SplitpointsBufferId,
random_splitpoint_selection_step.SplitpointsCountsBufferId,
two_stream_split_stats_step.ChildCountsEstimatorBufferId,
two_stream_split_stats_step.LeftEstimatorStatsBufferId,
two_stream_split_stats_step.RightEstimatorStatsBufferId,
feature_params_step.FloatParamsBufferId,
feature_params_step.IntParamsBufferId,
matrix_feature_extractor_step.FeatureValuesBufferId,
feature_ordering,
matrix_feature_extractor_step,
)
should_split_criteria = create_should_split_criteria(**kwargs)
finalizer = classification.ClassEstimatorFinalizer_f32()
split_indices = splitpoints.SplitIndices_f32i32(sample_data_step.IndicesBufferId)
split_selector = splitpoints.SplitSelector_f32i32([split_buffers], should_split_criteria, finalizer, split_indices)
if "tree_order" in kwargs and kwargs.get("tree_order") == "breadth_first":
tree_learner = learn.BreadthFirstTreeLearner_f32i32(
try_split_criteria, tree_steps_pipeline, node_steps_pipeline, split_selector
)
else:
tree_learner = learn.DepthFirstTreeLearner_f32i32(
try_split_criteria, tree_steps_pipeline, node_steps_pipeline, split_selector
)
forest_learner = learn.ParallelForestLearner(tree_learner, number_of_trees, number_of_classes, number_of_jobs)
return forest_learner
def create_biau2008_regression_scaled_depth_delta_learner_32f(**kwargs):
ux = float(kwargs.get('ux'))
uy = float(kwargs.get('uy'))
vx = float(kwargs.get('vx'))
vy = float(kwargs.get('vy'))
number_of_trees = int(kwargs.get('number_of_trees', 10))
number_of_leaves = int(
kwargs.get('number_of_leaves', kwargs['y'].GetM() / 5 + 1))
number_of_split_retries = int(kwargs.get('number_of_split_retries', 10))
number_of_features = 1
feature_ordering = int(
kwargs.get('feature_ordering', pipeline.FEATURES_BY_DATAPOINTS))
number_of_jobs = int(kwargs.get('number_of_jobs', 1))
dimension_of_y = int(kwargs['y'].GetN())
try_split_criteria = try_split.MinNodeSizeCriteria(2)
sample_data_step = pipeline.AllSamplesStep_i32f32i32(buffers.PIXEL_INDICES)
number_of_features_buffer = buffers.as_vector_buffer(
np.array([number_of_features], dtype=np.int32))
set_number_features_step = pipeline.SetInt32VectorBufferStep(
number_of_features_buffer, pipeline.WHEN_NEW)
tree_steps_pipeline = pipeline.Pipeline(
[sample_data_step, set_number_features_step])
feature_params_step = image_features.PixelPairGaussianOffsetsStep_f32i32(
set_number_features_step.OutputBufferId, ux, uy, vx, vy)
depth_delta_feature = image_features.ScaledDepthDeltaFeature_f32i32(
feature_params_step.FloatParamsBufferId,
feature_params_step.IntParamsBufferId,
sample_data_step.IndicesBufferId, buffers.PIXEL_INDICES,
buffers.DEPTH_IMAGES, buffers.OFFSET_SCALES)
depth_delta_feature_extractor_step = image_features.ScaledDepthDeltaFeatureExtractorStep_f32i32(
depth_delta_feature, feature_ordering)
slice_ys_step = pipeline.SliceFloat32MatrixBufferStep_i32(
buffers.YS, sample_data_step.IndicesBufferId)
slice_weights_step = pipeline.SliceFloat32VectorBufferStep_i32(
sample_data_step.WeightsBufferId, sample_data_step.IndicesBufferId)
impurity_walker = regression.SumOfVarianceWalker_f32i32(
slice_weights_step.SlicedBufferId, slice_ys_step.SlicedBufferId,
dimension_of_y)
best_splitpint_step = regression.SumOfVarianceRandomGapSplitpointsStep_f32i32(
impurity_walker,
depth_delta_feature_extractor_step.FeatureValuesBufferId,
feature_ordering)
node_steps_pipeline = pipeline.Pipeline([
feature_params_step, depth_delta_feature_extractor_step, slice_ys_step,
slice_weights_step, best_splitpint_step
])
split_buffers = splitpoints.SplitSelectorBuffers(
best_splitpint_step.ImpurityBufferId,
best_splitpint_step.SplitpointBufferId,
best_splitpint_step.SplitpointCountsBufferId,
best_splitpint_step.ChildCountsBufferId,
best_splitpint_step.LeftYsBufferId,
best_splitpint_step.RightYsBufferId,
feature_params_step.FloatParamsBufferId,
feature_params_step.IntParamsBufferId,
depth_delta_feature_extractor_step.FeatureValuesBufferId,
feature_ordering, depth_delta_feature_extractor_step)
should_split_criteria = min_child_size_criteria = should_split.MinChildSizeCriteria(
1)
finalizer = regression.MeanVarianceEstimatorFinalizer_f32()
split_indices = splitpoints.SplitIndices_f32i32(
sample_data_step.IndicesBufferId)
split_selector = splitpoints.SplitSelector_f32i32([split_buffers],
should_split_criteria,
finalizer, split_indices)
tree_learner = learn.Biau2008TreeLearner_f32i32(
try_split_criteria, tree_steps_pipeline, node_steps_pipeline,
split_selector, number_of_leaves, number_of_split_retries)
forest_learner = learn.ParallelForestLearner(tree_learner, number_of_trees,
dimension_of_y,
number_of_jobs)
return forest_learner
def create_regression_axis_aligned_matrix_learner_32f(**kwargs):
number_of_trees = int(kwargs.get('number_of_trees', 10))
number_of_leaves = int(
kwargs.get('number_of_leaves', kwargs['y'].shape[0] / 5 + 1))
number_of_features = int(
kwargs.get('number_of_features', (kwargs['x'].shape[1]) / 3 + 0.5))
# number_of_features = int( kwargs.get('number_of_features', np.sqrt(kwargs['x'].shape[1])))
feature_ordering = int(
kwargs.get('feature_ordering', pipeline.FEATURES_BY_DATAPOINTS))
number_of_jobs = int(kwargs.get('number_of_jobs', 1))
dimension_of_y = int(kwargs['y'].shape[1])
try_split_criteria = create_try_split_criteria(**kwargs)
if 'bootstrap' in kwargs and kwargs.get('bootstrap'):
sample_data_step = pipeline.BootstrapSamplesStep_f32f32i32(
buffers.X_FLOAT_DATA)
else:
sample_data_step = pipeline.AllSamplesStep_f32f32i32(
buffers.X_FLOAT_DATA)
number_of_features_buffer = buffers.as_vector_buffer(
np.array([number_of_features], dtype=np.int32))
set_number_features_step = pipeline.SetInt32VectorBufferStep(
number_of_features_buffer, pipeline.WHEN_NEW)
tree_steps_pipeline = pipeline.Pipeline(
[sample_data_step, set_number_features_step])
feature_params_step = matrix_features.AxisAlignedParamsStep_f32i32(
set_number_features_step.OutputBufferId, buffers.X_FLOAT_DATA)
matrix_feature = matrix_features.LinearFloat32MatrixFeature_f32i32(
feature_params_step.FloatParamsBufferId,
feature_params_step.IntParamsBufferId,
sample_data_step.IndicesBufferId, buffers.X_FLOAT_DATA)
matrix_feature_extractor_step = matrix_features.LinearFloat32MatrixFeatureExtractorStep_f32i32(
matrix_feature, feature_ordering)
slice_ys_step = pipeline.SliceFloat32MatrixBufferStep_i32(
buffers.YS, sample_data_step.IndicesBufferId)
slice_weights_step = pipeline.SliceFloat32VectorBufferStep_i32(
sample_data_step.WeightsBufferId, sample_data_step.IndicesBufferId)
impurity_walker = regression.SumOfVarianceWalker_f32i32(
slice_weights_step.SlicedBufferId, slice_ys_step.SlicedBufferId,
dimension_of_y)
best_splitpint_step = regression.SumOfVarianceBestSplitpointsWalkingSortedStep_f32i32(
impurity_walker, matrix_feature_extractor_step.FeatureValuesBufferId,
feature_ordering)
node_steps_pipeline = pipeline.Pipeline([
feature_params_step, matrix_feature_extractor_step, slice_ys_step,
slice_weights_step, best_splitpint_step
])
split_buffers = splitpoints.SplitSelectorBuffers(
best_splitpint_step.ImpurityBufferId,
best_splitpint_step.SplitpointBufferId,
best_splitpint_step.SplitpointCountsBufferId,
best_splitpint_step.ChildCountsBufferId,
best_splitpint_step.LeftYsBufferId,
best_splitpint_step.RightYsBufferId,
feature_params_step.FloatParamsBufferId,
feature_params_step.IntParamsBufferId,
matrix_feature_extractor_step.FeatureValuesBufferId, feature_ordering,
matrix_feature_extractor_step)
should_split_criteria = create_should_split_criteria(**kwargs)
finalizer = regression.MeanVarianceEstimatorFinalizer_f32()
split_indices = splitpoints.SplitIndices_f32i32(
sample_data_step.IndicesBufferId)
split_selector = splitpoints.SplitSelector_f32i32([split_buffers],
should_split_criteria,
finalizer, split_indices)
tree_learner = learn.BreadthFirstTreeLearner_f32i32(
try_split_criteria, tree_steps_pipeline, node_steps_pipeline,
split_selector, number_of_leaves)
forest_learner = learn.ParallelForestLearner(tree_learner, number_of_trees,
dimension_of_y,
number_of_jobs)
return forest_learner
def create_biau2012_regression_axis_aligned_matrix_learner_32f(**kwargs):
number_of_trees = int( kwargs.get('number_of_trees', 10) )
number_of_leaves = int( kwargs.get('number_of_leaves', kwargs['y'].shape[0] / 5 + 1) )
number_of_features = int( kwargs.get('number_of_features', (kwargs['x'].shape[1])/3 + 0.5))
# number_of_features = int( kwargs.get('number_of_features', np.sqrt(kwargs['x'].shape[1])))
feature_ordering = int( kwargs.get('feature_ordering', pipeline.FEATURES_BY_DATAPOINTS) )
number_of_jobs = int( kwargs.get('number_of_jobs', 1) )
dimension_of_y = int( kwargs['y'].shape[1] )
probability_of_impurity_stream = float(kwargs.get('probability_of_impurity_stream', 0.5) )
try_split_criteria = create_try_split_criteria(**kwargs)
sample_data_step = pipeline.AllSamplesStep_f32f32i32(buffers.X_FLOAT_DATA)
number_of_features_buffer = buffers.as_vector_buffer(np.array([number_of_features], dtype=np.int32))
set_number_features_step = pipeline.SetInt32VectorBufferStep(number_of_features_buffer, pipeline.WHEN_NEW)
assert(np.max(np.abs(kwargs['x'])) <= 1.00) # double check that the data has been scaled into a -1,1 hypercube
feature_range_buffer = buffers.as_vector_buffer(np.array([-1, 1], dtype=np.float32))
set_feature_range_buffer_step = pipeline.SetFloat32VectorBufferStep(feature_range_buffer, pipeline.WHEN_NEW)
assign_stream_step = splitpoints.AssignStreamStep_f32i32(sample_data_step.WeightsBufferId, probability_of_impurity_stream, False)
forest_steps_pipeline = pipeline.Pipeline([sample_data_step, set_number_features_step, set_feature_range_buffer_step, assign_stream_step])
tree_steps_pipeline = pipeline.Pipeline([])
feature_params_step = matrix_features.AxisAlignedParamsStep_f32i32(set_number_features_step.OutputBufferId, buffers.X_FLOAT_DATA)
matrix_feature = matrix_features.LinearFloat32MatrixFeature_f32i32(feature_params_step.FloatParamsBufferId,
feature_params_step.IntParamsBufferId,
sample_data_step.IndicesBufferId,
buffers.X_FLOAT_DATA)
matrix_feature_extractor_step = matrix_features.LinearFloat32MatrixFeatureExtractorStep_f32i32(matrix_feature, feature_ordering)
slice_ys_step = pipeline.SliceFloat32MatrixBufferStep_i32(buffers.YS, sample_data_step.IndicesBufferId)
slice_weights_step = pipeline.SliceFloat32VectorBufferStep_i32(sample_data_step.WeightsBufferId, sample_data_step.IndicesBufferId)
slice_assign_stream_step = pipeline.SliceInt32VectorBufferStep_i32(assign_stream_step.StreamTypeBufferId, sample_data_step.IndicesBufferId)
quantized_feature_equal = pipeline.FeatureEqualQuantized_f32i32(1.0)
midpoint_step = splitpoints.RangeMidpointStep_f32i32(feature_params_step.FloatParamsBufferId,
feature_params_step.IntParamsBufferId,
set_feature_range_buffer_step.OutputBufferId,
quantized_feature_equal)
mean_variance_stats_updater = regression.MeanVarianceStatsUpdater_f32i32(slice_weights_step.SlicedBufferId,
slice_ys_step.SlicedBufferId,
dimension_of_y)
two_stream_split_stats_step = regression.SumOfVarianceTwoStreamStep_f32i32(midpoint_step.SplitpointsBufferId,
midpoint_step.SplitpointsCountsBufferId,
slice_assign_stream_step.SlicedBufferId,
matrix_feature_extractor_step.FeatureValuesBufferId,
feature_ordering,
mean_variance_stats_updater)
impurity_step = regression.SumOfVarianceSplitpointsImpurity_f32i32(midpoint_step.SplitpointsCountsBufferId,
two_stream_split_stats_step.ChildCountsImpurityBufferId,
two_stream_split_stats_step.LeftImpurityStatsBufferId,
two_stream_split_stats_step.RightImpurityStatsBufferId)
node_steps_pipeline = pipeline.Pipeline([feature_params_step, matrix_feature_extractor_step,
slice_ys_step, slice_weights_step, slice_assign_stream_step,
midpoint_step, two_stream_split_stats_step, impurity_step])
split_buffers = splitpoints.SplitSelectorBuffers(impurity_step.ImpurityBufferId,
midpoint_step.SplitpointsBufferId,
midpoint_step.SplitpointsCountsBufferId,
two_stream_split_stats_step.ChildCountsEstimatorBufferId,
two_stream_split_stats_step.LeftEstimatorStatsBufferId,
two_stream_split_stats_step.RightEstimatorStatsBufferId,
feature_params_step.FloatParamsBufferId,
feature_params_step.IntParamsBufferId,
matrix_feature_extractor_step.FeatureValuesBufferId,
feature_ordering,
matrix_feature_extractor_step)
should_split_criteria = no_split_criteria(**kwargs)
finalizer = regression.MeanVarianceEstimatorFinalizer_f32()
split_indices = splitpoints.SplitIndices_f32i32(sample_data_step.IndicesBufferId)
split_midpoint_ranges = splitpoints.SplitBuffersFeatureRange_f32i32(midpoint_step.PastFloatParamsBufferId,
midpoint_step.PastIntParamsBufferId,
midpoint_step.PastRangesBufferId,
set_feature_range_buffer_step.OutputBufferId,
quantized_feature_equal)
split_steps = splitpoints.SplitBuffersList([split_indices, split_midpoint_ranges])
split_selector = splitpoints.SplitSelector_f32i32([split_buffers],
should_split_criteria,
finalizer,
split_steps)
tree_learner = learn.BreadthFirstTreeLearner_f32i32(try_split_criteria, tree_steps_pipeline, node_steps_pipeline, split_selector, number_of_leaves)
forest_learner = learn.ParallelForestLearner(tree_learner, forest_steps_pipeline, number_of_trees, dimension_of_y, number_of_jobs)
return forest_learner
# Load single pose depth and class labels
depths = kinect_utils.load_depth_from_exr("%s%s.exr" % (args.pose_files_input_path, pose_filename))
labels = kinect_utils.load_labels_from_png("%s%s.png" % (args.pose_files_input_path, pose_filename))
pixel_indices, pixel_labels = kinect_utils.sample_pixels_from_image(labels[0,:,:], config.number_of_pixels_per_image)
# Randomly sample pixels and offset scales
(number_of_datapoints, _) = pixel_indices.shape
offset_scales = np.array(np.random.uniform(0.8, 1.2, (number_of_datapoints, 2)), dtype=np.float32)
datapoint_indices = np.array(np.arange(number_of_datapoints), dtype=np.int32)
# Package buffers for learner
bufferCollection = buffers.BufferCollection()
bufferCollection.AddFloat32Tensor3Buffer(buffers.DEPTH_IMAGES, buffers.as_tensor_buffer(depths))
bufferCollection.AddFloat32MatrixBuffer(buffers.OFFSET_SCALES, buffers.as_matrix_buffer(offset_scales))
bufferCollection.AddInt32MatrixBuffer(buffers.PIXEL_INDICES, buffers.as_matrix_buffer(pixel_indices))
bufferCollection.AddInt32VectorBuffer(buffers.CLASS_LABELS, buffers.as_vector_buffer(pixel_labels))
# Update learner
online_learner.Train(bufferCollection, buffers.Int32Vector(datapoint_indices))
#pickle forest and data used for training
if (i+1) % 1 == 0:
# if True:
forest_pickle_filename = "%s/forest-%d-%d.pkl" % (online_run_folder, pass_id, i+1)
pickle.dump(online_learner.GetForest(), gzip.open(forest_pickle_filename, 'wb'))
# Print forest stats
forestStats = online_learner.GetForest().GetForestStats()
forestStats.Print()
offset_scales = np.array(np.random.uniform(
0.8, 1.2, (number_of_datapoints, 2)),
dtype=np.float32)
datapoint_indices = np.array(np.arange(number_of_datapoints),
dtype=np.int32)
# Package buffers for learner
bufferCollection = buffers.BufferCollection()
bufferCollection.AddFloat32Tensor3Buffer(
buffers.DEPTH_IMAGES, buffers.as_tensor_buffer(depths))
bufferCollection.AddFloat32MatrixBuffer(
buffers.OFFSET_SCALES, buffers.as_matrix_buffer(offset_scales))
bufferCollection.AddInt32MatrixBuffer(
buffers.PIXEL_INDICES, buffers.as_matrix_buffer(pixel_indices))
bufferCollection.AddInt32VectorBuffer(
buffers.CLASS_LABELS, buffers.as_vector_buffer(pixel_labels))
# Update learner
online_learner.Train(bufferCollection,
buffers.Int32Vector(datapoint_indices))
#pickle forest and data used for training
if (i + 1) % 1 == 0:
# if True:
forest_pickle_filename = "%s/forest-%d-%d.pkl" % (
online_run_folder, pass_id, i + 1)
pickle.dump(online_learner.GetForest(),
gzip.open(forest_pickle_filename, 'wb'))
# Print forest stats
forestStats = online_learner.GetForest().GetForestStats()
def create_biau2012_regression_scaled_depth_delta_learner_32f(**kwargs):
ux = float(kwargs.get('ux'))
uy = float(kwargs.get('uy'))
vx = float(kwargs.get('vx'))
vy = float(kwargs.get('vy'))
number_of_trees = int(kwargs.get('number_of_trees', 10))
number_of_leaves = int(
kwargs.get('number_of_leaves', kwargs['y'].GetM() / 5 + 1))
number_of_features = int(kwargs.get('number_of_features', 1))
feature_ordering = int(
kwargs.get('feature_ordering', pipeline.FEATURES_BY_DATAPOINTS))
number_of_jobs = int(kwargs.get('number_of_jobs', 1))
dimension_of_y = int(kwargs['y'].GetN())
probability_of_impurity_stream = float(
kwargs.get('probability_of_impurity_stream', 0.5))
try_split_criteria = create_try_split_criteria(**kwargs)
sample_data_step = pipeline.AllSamplesStep_i32f32i32(buffers.PIXEL_INDICES)
number_of_features_buffer = buffers.as_vector_buffer(
np.array([number_of_features], dtype=np.int32))
set_number_features_step = pipeline.SetInt32VectorBufferStep(
number_of_features_buffer, pipeline.WHEN_NEW)
feature_range_buffer = buffers.as_vector_buffer(
np.array([-6, 6], dtype=np.float32))
set_feature_range_buffer_step = pipeline.SetFloat32VectorBufferStep(
feature_range_buffer, pipeline.WHEN_NEW)
assign_stream_step = splitpoints.AssignStreamStep_f32i32(
sample_data_step.WeightsBufferId, probability_of_impurity_stream,
False)
forest_steps_pipeline = pipeline.Pipeline([
sample_data_step, set_number_features_step,
set_feature_range_buffer_step, assign_stream_step
])
tree_steps_pipeline = pipeline.Pipeline([])
feature_params_step = image_features.PixelPairGaussianOffsetsStep_f32i32(
set_number_features_step.OutputBufferId, ux, uy, vx, vy)
depth_delta_feature = image_features.ScaledDepthDeltaFeature_f32i32(
feature_params_step.FloatParamsBufferId,
feature_params_step.IntParamsBufferId,
sample_data_step.IndicesBufferId, buffers.PIXEL_INDICES,
buffers.DEPTH_IMAGES, buffers.OFFSET_SCALES)
depth_delta_feature_extractor_step = image_features.ScaledDepthDeltaFeatureExtractorStep_f32i32(
depth_delta_feature, feature_ordering)
slice_ys_step = pipeline.SliceFloat32MatrixBufferStep_i32(
buffers.YS, sample_data_step.IndicesBufferId)
slice_weights_step = pipeline.SliceFloat32VectorBufferStep_i32(
sample_data_step.WeightsBufferId, sample_data_step.IndicesBufferId)
slice_assign_stream_step = pipeline.SliceInt32VectorBufferStep_i32(
assign_stream_step.StreamTypeBufferId,
sample_data_step.IndicesBufferId)
quantized_feature_equal = pipeline.FeatureEqualQuantized_f32i32(1.0)
midpoint_step = splitpoints.RangeMidpointStep_f32i32(
feature_params_step.FloatParamsBufferId,
feature_params_step.IntParamsBufferId,
set_feature_range_buffer_step.OutputBufferId, quantized_feature_equal)
mean_variance_stats_updater = regression.MeanVarianceStatsUpdater_f32i32(
slice_weights_step.SlicedBufferId, slice_ys_step.SlicedBufferId,
dimension_of_y)
two_stream_split_stats_step = regression.SumOfVarianceTwoStreamStep_f32i32(
midpoint_step.SplitpointsBufferId,
midpoint_step.SplitpointsCountsBufferId,
slice_assign_stream_step.SlicedBufferId,
depth_delta_feature_extractor_step.FeatureValuesBufferId,
feature_ordering, mean_variance_stats_updater)
impurity_step = regression.SumOfVarianceSplitpointsImpurity_f32i32(
midpoint_step.SplitpointsCountsBufferId,
two_stream_split_stats_step.ChildCountsImpurityBufferId,
two_stream_split_stats_step.LeftImpurityStatsBufferId,
two_stream_split_stats_step.RightImpurityStatsBufferId)
node_steps_pipeline = pipeline.Pipeline([
feature_params_step, depth_delta_feature_extractor_step, slice_ys_step,
slice_weights_step, slice_assign_stream_step, midpoint_step,
two_stream_split_stats_step, impurity_step
])
split_buffers = splitpoints.SplitSelectorBuffers(
impurity_step.ImpurityBufferId, midpoint_step.SplitpointsBufferId,
midpoint_step.SplitpointsCountsBufferId,
two_stream_split_stats_step.ChildCountsEstimatorBufferId,
two_stream_split_stats_step.LeftEstimatorStatsBufferId,
two_stream_split_stats_step.RightEstimatorStatsBufferId,
feature_params_step.FloatParamsBufferId,
feature_params_step.IntParamsBufferId,
depth_delta_feature_extractor_step.FeatureValuesBufferId,
feature_ordering, depth_delta_feature_extractor_step)
should_split_criteria = no_split_criteria(**kwargs)
finalizer = regression.MeanVarianceEstimatorFinalizer_f32()
split_indices = splitpoints.SplitIndices_f32i32(
sample_data_step.IndicesBufferId)
split_midpoint_ranges = splitpoints.SplitBuffersFeatureRange_f32i32(
midpoint_step.PastFloatParamsBufferId,
midpoint_step.PastIntParamsBufferId, midpoint_step.PastRangesBufferId,
set_feature_range_buffer_step.OutputBufferId, quantized_feature_equal)
split_steps = splitpoints.SplitBuffersList(
[split_indices, split_midpoint_ranges])
split_selector = splitpoints.SplitSelector_f32i32([split_buffers],
should_split_criteria,
finalizer, split_steps)
tree_learner = learn.BreadthFirstTreeLearner_f32i32(
try_split_criteria, tree_steps_pipeline, node_steps_pipeline,
split_selector, number_of_leaves)
forest_learner = learn.ParallelForestLearner(tree_learner,
forest_steps_pipeline,
number_of_trees,
dimension_of_y,
number_of_jobs)
return forest_learner
def create_online_axis_aligned_matrix_one_stream_learner_32f(**kwargs):
number_of_trees = int(kwargs.get('number_of_trees', 10))
number_of_features = int(
kwargs.get('number_of_features', np.sqrt(kwargs['x'].shape[1])))
feature_ordering = int(
kwargs.get('feature_ordering', pipeline.FEATURES_BY_DATAPOINTS))
number_of_splitpoints = int(kwargs.get('number_of_splitpoints', 1))
number_of_classes = int(np.max(kwargs['classes']) + 1)
max_frontier_size = int(kwargs.get('max_frontier_size', 10000000))
impurity_update_period = int(kwargs.get('impurity_update_period', 1))
try_split_criteria = create_try_split_criteria(**kwargs)
if 'bootstrap' in kwargs and kwargs.get('bootstrap'):
sample_data_step = pipeline.BootstrapSamplesStep_f32f32i32(
buffers.X_FLOAT_DATA)
elif 'poisson_sample' in kwargs:
poisson_sample_mean = float(kwargs.get('poisson_sample'))
sample_data_step = pipeline.PoissonSamplesStep_f32i32(
buffers.X_FLOAT_DATA, poisson_sample_mean)
else:
sample_data_step = pipeline.AllSamplesStep_f32f32i32(
buffers.X_FLOAT_DATA)
number_of_features_buffer = buffers.as_vector_buffer(
np.array([number_of_features], dtype=np.int32))
set_number_features_step = pipeline.SetInt32VectorBufferStep(
number_of_features_buffer, pipeline.WHEN_NEW)
tree_steps_pipeline = pipeline.Pipeline(
[sample_data_step, set_number_features_step])
feature_params_step = matrix_features.AxisAlignedParamsStep_f32i32(
set_number_features_step.OutputBufferId, buffers.X_FLOAT_DATA)
matrix_feature = matrix_features.LinearFloat32MatrixFeature_f32i32(
feature_params_step.FloatParamsBufferId,
feature_params_step.IntParamsBufferId,
sample_data_step.IndicesBufferId, buffers.X_FLOAT_DATA)
matrix_feature_extractor_step = matrix_features.LinearFloat32MatrixFeatureExtractorStep_f32i32(
matrix_feature, feature_ordering)
slice_classes_step = pipeline.SliceInt32VectorBufferStep_i32(
buffers.CLASS_LABELS, sample_data_step.IndicesBufferId)
slice_weights_step = pipeline.SliceFloat32VectorBufferStep_i32(
sample_data_step.WeightsBufferId, sample_data_step.IndicesBufferId)
random_splitpoint_selection_step = splitpoints.RandomSplitpointsStep_f32i32(
matrix_feature_extractor_step.FeatureValuesBufferId,
number_of_splitpoints, feature_ordering)
class_stats_updater = classification.ClassStatsUpdater_f32i32(
slice_weights_step.SlicedBufferId, slice_classes_step.SlicedBufferId,
number_of_classes)
one_stream_split_stats_step = classification.ClassStatsUpdaterOneStreamStep_f32i32(
random_splitpoint_selection_step.SplitpointsBufferId,
random_splitpoint_selection_step.SplitpointsCountsBufferId,
matrix_feature_extractor_step.FeatureValuesBufferId, feature_ordering,
class_stats_updater)
impurity_step = classification.ClassInfoGainSplitpointsImpurity_f32i32(
random_splitpoint_selection_step.SplitpointsCountsBufferId,
one_stream_split_stats_step.ChildCountsBufferId,
one_stream_split_stats_step.LeftStatsBufferId,
one_stream_split_stats_step.RightStatsBufferId)
init_node_steps_pipeline = pipeline.Pipeline([feature_params_step])
update_stats_node_steps_pipeline = pipeline.Pipeline([
matrix_feature_extractor_step, slice_classes_step, slice_weights_step,
random_splitpoint_selection_step, one_stream_split_stats_step
])
update_impurity_node_steps_pipeline = pipeline.Pipeline([impurity_step])
split_buffers = splitpoints.SplitSelectorBuffers(
impurity_step.ImpurityBufferId,
random_splitpoint_selection_step.SplitpointsBufferId,
random_splitpoint_selection_step.SplitpointsCountsBufferId,
one_stream_split_stats_step.ChildCountsBufferId,
one_stream_split_stats_step.LeftStatsBufferId,
one_stream_split_stats_step.RightStatsBufferId,
feature_params_step.FloatParamsBufferId,
feature_params_step.IntParamsBufferId,
matrix_feature_extractor_step.FeatureValuesBufferId, feature_ordering,
matrix_feature_extractor_step)
should_split_criteria = create_should_split_criteria(**kwargs)
finalizer = classification.ClassEstimatorFinalizer_f32()
split_indices = splitpoints.SplitIndices_f32i32(
sample_data_step.IndicesBufferId)
split_selector = splitpoints.WaitForBestSplitSelector_f32i32(
[split_buffers], should_split_criteria, finalizer, split_indices)
matrix_feature_prediction = matrix_features.LinearFloat32MatrixFeature_f32i32(
sample_data_step.IndicesBufferId, buffers.X_FLOAT_DATA)
estimator_params_updater = classification.ClassEstimatorUpdater_f32i32(
sample_data_step.WeightsBufferId, buffers.CLASS_LABELS,
number_of_classes)
forest_learner = learn.OnlineForestMatrixClassLearner_f32i32(
try_split_criteria, tree_steps_pipeline, init_node_steps_pipeline,
update_stats_node_steps_pipeline, update_impurity_node_steps_pipeline,
impurity_update_period, split_selector, max_frontier_size,
number_of_trees, 5, 5, number_of_classes,
sample_data_step.IndicesBufferId, sample_data_step.WeightsBufferId,
matrix_feature_prediction, estimator_params_updater)
return forest_learner
def create_biau2012_regression_scaled_depth_delta_learner_32f(**kwargs):
ux = float( kwargs.get('ux') )
uy = float( kwargs.get('uy') )
vx = float( kwargs.get('vx') )
vy = float( kwargs.get('vy') )
number_of_trees = int( kwargs.get('number_of_trees', 10) )
number_of_leaves = int( kwargs.get('number_of_leaves', kwargs['y'].GetM() / 5 + 1) )
number_of_features = int( kwargs.get('number_of_features', 1) )
feature_ordering = int( kwargs.get('feature_ordering', pipeline.FEATURES_BY_DATAPOINTS) )
number_of_jobs = int( kwargs.get('number_of_jobs', 1) )
dimension_of_y = int( kwargs['y'].GetN() )
probability_of_impurity_stream = float(kwargs.get('probability_of_impurity_stream', 0.5) )
try_split_criteria = create_try_split_criteria(**kwargs)
sample_data_step = pipeline.AllSamplesStep_i32f32i32(buffers.PIXEL_INDICES)
number_of_features_buffer = buffers.as_vector_buffer(np.array([number_of_features], dtype=np.int32))
set_number_features_step = pipeline.SetInt32VectorBufferStep(number_of_features_buffer, pipeline.WHEN_NEW)
feature_range_buffer = buffers.as_vector_buffer(np.array([-6, 6], dtype=np.float32))
set_feature_range_buffer_step = pipeline.SetFloat32VectorBufferStep(feature_range_buffer, pipeline.WHEN_NEW)
assign_stream_step = splitpoints.AssignStreamStep_f32i32(sample_data_step.WeightsBufferId, probability_of_impurity_stream, False)
forest_steps_pipeline = pipeline.Pipeline([sample_data_step, set_number_features_step, set_feature_range_buffer_step, assign_stream_step])
tree_steps_pipeline = pipeline.Pipeline([])
feature_params_step = image_features.PixelPairGaussianOffsetsStep_f32i32(set_number_features_step.OutputBufferId, ux, uy, vx, vy )
depth_delta_feature = image_features.ScaledDepthDeltaFeature_f32i32(feature_params_step.FloatParamsBufferId,
feature_params_step.IntParamsBufferId,
sample_data_step.IndicesBufferId,
buffers.PIXEL_INDICES,
buffers.DEPTH_IMAGES,
buffers.OFFSET_SCALES)
depth_delta_feature_extractor_step = image_features.ScaledDepthDeltaFeatureExtractorStep_f32i32(depth_delta_feature, feature_ordering)
slice_ys_step = pipeline.SliceFloat32MatrixBufferStep_i32(buffers.YS, sample_data_step.IndicesBufferId)
slice_weights_step = pipeline.SliceFloat32VectorBufferStep_i32(sample_data_step.WeightsBufferId, sample_data_step.IndicesBufferId)
slice_assign_stream_step = pipeline.SliceInt32VectorBufferStep_i32(assign_stream_step.StreamTypeBufferId, sample_data_step.IndicesBufferId)
quantized_feature_equal = pipeline.FeatureEqualQuantized_f32i32(1.0)
midpoint_step = splitpoints.RangeMidpointStep_f32i32(feature_params_step.FloatParamsBufferId,
feature_params_step.IntParamsBufferId,
set_feature_range_buffer_step.OutputBufferId,
quantized_feature_equal)
mean_variance_stats_updater = regression.MeanVarianceStatsUpdater_f32i32(slice_weights_step.SlicedBufferId,
slice_ys_step.SlicedBufferId,
dimension_of_y)
two_stream_split_stats_step = regression.SumOfVarianceTwoStreamStep_f32i32(midpoint_step.SplitpointsBufferId,
midpoint_step.SplitpointsCountsBufferId,
slice_assign_stream_step.SlicedBufferId,
depth_delta_feature_extractor_step.FeatureValuesBufferId,
feature_ordering,
mean_variance_stats_updater)
impurity_step = regression.SumOfVarianceSplitpointsImpurity_f32i32(midpoint_step.SplitpointsCountsBufferId,
two_stream_split_stats_step.ChildCountsImpurityBufferId,
two_stream_split_stats_step.LeftImpurityStatsBufferId,
two_stream_split_stats_step.RightImpurityStatsBufferId)
node_steps_pipeline = pipeline.Pipeline([feature_params_step, depth_delta_feature_extractor_step,
slice_ys_step, slice_weights_step, slice_assign_stream_step,
midpoint_step, two_stream_split_stats_step, impurity_step])
split_buffers = splitpoints.SplitSelectorBuffers(impurity_step.ImpurityBufferId,
midpoint_step.SplitpointsBufferId,
midpoint_step.SplitpointsCountsBufferId,
two_stream_split_stats_step.ChildCountsEstimatorBufferId,
two_stream_split_stats_step.LeftEstimatorStatsBufferId,
two_stream_split_stats_step.RightEstimatorStatsBufferId,
feature_params_step.FloatParamsBufferId,
feature_params_step.IntParamsBufferId,
depth_delta_feature_extractor_step.FeatureValuesBufferId,
feature_ordering,
depth_delta_feature_extractor_step)
should_split_criteria = no_split_criteria(**kwargs)
finalizer = regression.MeanVarianceEstimatorFinalizer_f32()
split_indices = splitpoints.SplitIndices_f32i32(sample_data_step.IndicesBufferId)
split_midpoint_ranges = splitpoints.SplitBuffersFeatureRange_f32i32(midpoint_step.PastFloatParamsBufferId,
midpoint_step.PastIntParamsBufferId,
midpoint_step.PastRangesBufferId,
set_feature_range_buffer_step.OutputBufferId,
quantized_feature_equal)
split_steps = splitpoints.SplitBuffersList([split_indices, split_midpoint_ranges])
split_selector = splitpoints.SplitSelector_f32i32([split_buffers],
should_split_criteria,
finalizer,
split_steps)
tree_learner = learn.BreadthFirstTreeLearner_f32i32(try_split_criteria, tree_steps_pipeline, node_steps_pipeline, split_selector, number_of_leaves)
forest_learner = learn.ParallelForestLearner(tree_learner, forest_steps_pipeline, number_of_trees, dimension_of_y, number_of_jobs)
return forest_learner
def create_online_scaled_depth_delta_one_stream_learner_32f(**kwargs):
number_of_trees = int(kwargs.get('number_of_trees', 10))
number_of_features = int(kwargs.get('number_of_features', 1))
feature_ordering = int(
kwargs.get('feature_ordering', pipeline.FEATURES_BY_DATAPOINTS))
number_of_splitpoints = int(kwargs.get('number_of_splitpoints', 1))
number_of_classes = int(kwargs['classes'].GetMax() + 1)
max_frontier_size = int(kwargs.get('max_frontier_size', 10000000))
impurity_update_period = int(kwargs.get('impurity_update_period', 1))
ux = float(kwargs.get('ux'))
uy = float(kwargs.get('uy'))
vx = float(kwargs.get('vx'))
vy = float(kwargs.get('vy'))
try_split_criteria = create_try_split_criteria(**kwargs)
if 'bootstrap' in kwargs and kwargs.get('bootstrap'):
sample_data_step = pipeline.BootstrapSamplesStep_i32f32i32(
buffers.PIXEL_INDICES)
elif 'poisson_sample' in kwargs:
poisson_sample_mean = float(kwargs.get('poisson_sample'))
sample_data_step = pipeline.PoissonSamplesStep_i32i32(
buffers.PIXEL_INDICES, poisson_sample_mean)
else:
sample_data_step = pipeline.AllSamplesStep_i32f32i32(
buffers.PIXEL_INDICES)
number_of_features_buffer = buffers.as_vector_buffer(
np.array([number_of_features], dtype=np.int32))
set_number_features_step = pipeline.SetInt32VectorBufferStep(
number_of_features_buffer, pipeline.WHEN_NEW)
tree_steps_pipeline = pipeline.Pipeline(
[sample_data_step, set_number_features_step])
feature_params_step = image_features.PixelPairGaussianOffsetsStep_f32i32(
set_number_features_step.OutputBufferId, ux, uy, vx, vy)
init_node_steps_pipeline = pipeline.Pipeline([feature_params_step])
depth_delta_feature = image_features.ScaledDepthDeltaFeature_f32i32(
feature_params_step.FloatParamsBufferId,
feature_params_step.IntParamsBufferId,
sample_data_step.IndicesBufferId, buffers.PIXEL_INDICES,
buffers.DEPTH_IMAGES, buffers.OFFSET_SCALES)
depth_delta_feature_extractor_step = image_features.ScaledDepthDeltaFeatureExtractorStep_f32i32(
depth_delta_feature, feature_ordering)
slice_classes_step = pipeline.SliceInt32VectorBufferStep_i32(
buffers.CLASS_LABELS, sample_data_step.IndicesBufferId)
slice_weights_step = pipeline.SliceFloat32VectorBufferStep_i32(
sample_data_step.WeightsBufferId, sample_data_step.IndicesBufferId)
random_splitpoint_selection_step = splitpoints.RandomSplitpointsStep_f32i32(
depth_delta_feature_extractor_step.FeatureValuesBufferId,
number_of_splitpoints, feature_ordering)
class_stats_updater = classification.ClassStatsUpdater_f32i32(
slice_weights_step.SlicedBufferId, slice_classes_step.SlicedBufferId,
number_of_classes)
one_stream_split_stats_step = classification.ClassStatsUpdaterOneStreamStep_f32i32(
random_splitpoint_selection_step.SplitpointsBufferId,
random_splitpoint_selection_step.SplitpointsCountsBufferId,
depth_delta_feature_extractor_step.FeatureValuesBufferId,
feature_ordering, class_stats_updater)
update_stats_node_steps_pipeline = pipeline.Pipeline([
depth_delta_feature_extractor_step, slice_classes_step,
slice_weights_step, random_splitpoint_selection_step,
one_stream_split_stats_step
])
impurity_step = classification.ClassInfoGainSplitpointsImpurity_f32i32(
random_splitpoint_selection_step.SplitpointsCountsBufferId,
one_stream_split_stats_step.ChildCountsBufferId,
one_stream_split_stats_step.LeftStatsBufferId,
one_stream_split_stats_step.RightStatsBufferId)
update_impurity_node_steps_pipeline = pipeline.Pipeline([impurity_step])
split_buffers = splitpoints.SplitSelectorBuffers(
impurity_step.ImpurityBufferId,
random_splitpoint_selection_step.SplitpointsBufferId,
random_splitpoint_selection_step.SplitpointsCountsBufferId,
one_stream_split_stats_step.ChildCountsBufferId,
one_stream_split_stats_step.LeftStatsBufferId,
one_stream_split_stats_step.RightStatsBufferId,
feature_params_step.FloatParamsBufferId,
feature_params_step.IntParamsBufferId,
depth_delta_feature_extractor_step.FeatureValuesBufferId,
feature_ordering, depth_delta_feature_extractor_step)
should_split_criteria = create_should_split_criteria(**kwargs)
finalizer = classification.ClassEstimatorFinalizer_f32()
split_selector = splitpoints.WaitForBestSplitSelector_f32i32(
[split_buffers], should_split_criteria, finalizer)
feature_prediction = image_features.ScaledDepthDeltaFeature_f32i32(
sample_data_step.IndicesBufferId, buffers.PIXEL_INDICES,
buffers.DEPTH_IMAGES)
estimator_params_updater = classification.ClassEstimatorUpdater_f32i32(
sample_data_step.WeightsBufferId, buffers.CLASS_LABELS,
number_of_classes)
forest_learner = learn.OnlineForestScaledDepthDeltaClassLearner_f32i32(
try_split_criteria, tree_steps_pipeline, init_node_steps_pipeline,
update_stats_node_steps_pipeline, update_impurity_node_steps_pipeline,
impurity_update_period, split_selector, max_frontier_size,
number_of_trees, 5, 5, number_of_classes,
sample_data_step.IndicesBufferId, sample_data_step.WeightsBufferId,
feature_prediction, estimator_params_updater)
return forest_learner
def create_online_scaled_depth_delta_one_stream_learner_32f(**kwargs):
number_of_trees = int( kwargs.get('number_of_trees', 10) )
number_of_features = int( kwargs.get('number_of_features', 1))
feature_ordering = int( kwargs.get('feature_ordering', pipeline.FEATURES_BY_DATAPOINTS) )
number_of_splitpoints = int( kwargs.get('number_of_splitpoints', 1 ))
number_of_classes = int( kwargs['classes'].GetMax() + 1 )
max_frontier_size = int( kwargs.get('max_frontier_size', 10000000) )
impurity_update_period = int( kwargs.get('impurity_update_period', 1) )
ux = float( kwargs.get('ux') )
uy = float( kwargs.get('uy') )
vx = float( kwargs.get('vx') )
vy = float( kwargs.get('vy') )
try_split_criteria = create_try_split_criteria(**kwargs)
if 'bootstrap' in kwargs and kwargs.get('bootstrap'):
sample_data_step = pipeline.BootstrapSamplesStep_i32f32i32(buffers.PIXEL_INDICES)
elif 'poisson_sample' in kwargs:
poisson_sample_mean = float(kwargs.get('poisson_sample'))
sample_data_step = pipeline.PoissonSamplesStep_i32i32(buffers.PIXEL_INDICES, poisson_sample_mean)
else:
sample_data_step = pipeline.AllSamplesStep_i32f32i32(buffers.PIXEL_INDICES)
number_of_features_buffer = buffers.as_vector_buffer(np.array([number_of_features], dtype=np.int32))
set_number_features_step = pipeline.SetInt32VectorBufferStep(number_of_features_buffer, pipeline.WHEN_NEW)
tree_steps_pipeline = pipeline.Pipeline([sample_data_step, set_number_features_step])
feature_params_step = image_features.PixelPairGaussianOffsetsStep_f32i32(set_number_features_step.OutputBufferId, ux, uy, vx, vy )
init_node_steps_pipeline = pipeline.Pipeline([feature_params_step])
depth_delta_feature = image_features.ScaledDepthDeltaFeature_f32i32(feature_params_step.FloatParamsBufferId,
feature_params_step.IntParamsBufferId,
sample_data_step.IndicesBufferId,
buffers.PIXEL_INDICES,
buffers.DEPTH_IMAGES,
buffers.OFFSET_SCALES)
depth_delta_feature_extractor_step = image_features.ScaledDepthDeltaFeatureExtractorStep_f32i32(depth_delta_feature, feature_ordering)
slice_classes_step = pipeline.SliceInt32VectorBufferStep_i32(buffers.CLASS_LABELS, sample_data_step.IndicesBufferId)
slice_weights_step = pipeline.SliceFloat32VectorBufferStep_i32(sample_data_step.WeightsBufferId, sample_data_step.IndicesBufferId)
random_splitpoint_selection_step = splitpoints.RandomSplitpointsStep_f32i32(depth_delta_feature_extractor_step.FeatureValuesBufferId,
number_of_splitpoints,
feature_ordering)
class_stats_updater = classification.ClassStatsUpdater_f32i32(slice_weights_step.SlicedBufferId,
slice_classes_step.SlicedBufferId,
number_of_classes)
one_stream_split_stats_step = classification.ClassStatsUpdaterOneStreamStep_f32i32(random_splitpoint_selection_step.SplitpointsBufferId,
random_splitpoint_selection_step.SplitpointsCountsBufferId,
depth_delta_feature_extractor_step.FeatureValuesBufferId,
feature_ordering,
class_stats_updater)
update_stats_node_steps_pipeline = pipeline.Pipeline([depth_delta_feature_extractor_step,
slice_classes_step,
slice_weights_step,
random_splitpoint_selection_step,
one_stream_split_stats_step])
impurity_step = classification.ClassInfoGainSplitpointsImpurity_f32i32(random_splitpoint_selection_step.SplitpointsCountsBufferId,
one_stream_split_stats_step.ChildCountsBufferId,
one_stream_split_stats_step.LeftStatsBufferId,
one_stream_split_stats_step.RightStatsBufferId)
update_impurity_node_steps_pipeline = pipeline.Pipeline([impurity_step])
split_buffers = splitpoints.SplitSelectorBuffers(impurity_step.ImpurityBufferId,
random_splitpoint_selection_step.SplitpointsBufferId,
random_splitpoint_selection_step.SplitpointsCountsBufferId,
one_stream_split_stats_step.ChildCountsBufferId,
one_stream_split_stats_step.LeftStatsBufferId,
one_stream_split_stats_step.RightStatsBufferId,
feature_params_step.FloatParamsBufferId,
feature_params_step.IntParamsBufferId,
depth_delta_feature_extractor_step.FeatureValuesBufferId,
feature_ordering,
sample_data_step.IndicesBufferId)
should_split_criteria = create_should_split_criteria(**kwargs)
finalizer = classification.ClassEstimatorFinalizer_f32()
split_selector = splitpoints.WaitForBestSplitSelector_f32i32([split_buffers], should_split_criteria, finalizer )
feature_prediction = image_features.ScaledDepthDeltaFeature_f32i32(sample_data_step.IndicesBufferId,
buffers.PIXEL_INDICES,
buffers.DEPTH_IMAGES)
estimator_params_updater = classification.ClassEstimatorUpdater_f32i32(sample_data_step.WeightsBufferId, buffers.CLASS_LABELS, number_of_classes)
forest_learner = learn.OnlineForestScaledDepthDeltaClassLearner_f32i32(
try_split_criteria,
tree_steps_pipeline,
init_node_steps_pipeline,
update_stats_node_steps_pipeline,
update_impurity_node_steps_pipeline,
impurity_update_period, split_selector,
max_frontier_size, number_of_trees, 5, 5, number_of_classes,
sample_data_step.IndicesBufferId, sample_data_step.WeightsBufferId,
feature_prediction, estimator_params_updater)
return forest_learner
def create_consistent_two_stream_regression_axis_aligned_matrix_learner_32f(**kwargs):
number_of_trees = int( kwargs.get('number_of_trees', 10) )
number_of_leaves = int( kwargs.get('number_of_leaves', kwargs['y'].shape[0] / 5 + 1) )
number_of_features = int( kwargs.get('number_of_features', (kwargs['x'].shape[1])/3 + 0.5))
# number_of_features = int( kwargs.get('number_of_features', np.sqrt(kwargs['x'].shape[1])))
feature_ordering = int( kwargs.get('feature_ordering', pipeline.FEATURES_BY_DATAPOINTS) )
number_of_jobs = int( kwargs.get('number_of_jobs', 1) )
dimension_of_y = int( kwargs['y'].shape[1] )
probability_of_impurity_stream = float(kwargs.get('probability_of_impurity_stream', 0.5) )
in_bounds_number_of_points = int(kwargs.get('in_bounds_number_of_points', kwargs['y'].shape[0]/2) )
try_split_criteria = create_try_split_criteria(**kwargs)
if 'bootstrap' in kwargs and kwargs.get('bootstrap'):
sample_data_step = pipeline.BootstrapSamplesStep_f32f32i32(buffers.X_FLOAT_DATA)
else:
sample_data_step = pipeline.AllSamplesStep_f32f32i32(buffers.X_FLOAT_DATA)
assign_stream_step = splitpoints.AssignStreamStep_f32i32(sample_data_step.WeightsBufferId, probability_of_impurity_stream)
tree_steps_pipeline = pipeline.Pipeline([sample_data_step, assign_stream_step])
if 'poisson_number_of_features' in kwargs and kwargs.get('poisson_number_of_features'):
set_number_features_step = pipeline.PoissonStep_f32i32(number_of_features, 1)
else:
number_of_features_buffer = buffers.as_vector_buffer(np.array([number_of_features], dtype=np.int32))
set_number_features_step = pipeline.SetInt32VectorBufferStep(number_of_features_buffer, pipeline.WHEN_NEW)
feature_params_step = matrix_features.AxisAlignedParamsStep_f32i32(set_number_features_step.OutputBufferId, buffers.X_FLOAT_DATA)
matrix_feature = matrix_features.LinearFloat32MatrixFeature_f32i32(feature_params_step.FloatParamsBufferId,
feature_params_step.IntParamsBufferId,
sample_data_step.IndicesBufferId,
buffers.X_FLOAT_DATA)
matrix_feature_extractor_step = matrix_features.LinearFloat32MatrixFeatureExtractorStep_f32i32(matrix_feature, feature_ordering)
slice_ys_step = pipeline.SliceFloat32MatrixBufferStep_i32(buffers.YS, sample_data_step.IndicesBufferId)
slice_weights_step = pipeline.SliceFloat32VectorBufferStep_i32(sample_data_step.WeightsBufferId, sample_data_step.IndicesBufferId)
slice_stream_step = pipeline.SliceInt32VectorBufferStep_i32(assign_stream_step.StreamTypeBufferId, sample_data_step.IndicesBufferId)
impurity_walker = regression.SumOfVarianceTwoStreamWalker_f32i32(slice_weights_step.SlicedBufferId,
slice_stream_step.SlicedBufferId,
slice_ys_step.SlicedBufferId,
dimension_of_y)
best_splitpint_step = regression.SumOfVarianceTwoStreamBestSplitpointsWalkingSortedStep_f32i32(impurity_walker,
slice_stream_step.SlicedBufferId,
matrix_feature_extractor_step.FeatureValuesBufferId,
feature_ordering,
splitpoints.AT_MIDPOINT,
in_bounds_number_of_points)
node_steps_pipeline = pipeline.Pipeline([set_number_features_step, feature_params_step, matrix_feature_extractor_step,
slice_ys_step, slice_weights_step, slice_stream_step, best_splitpint_step])
split_buffers = splitpoints.SplitSelectorBuffers(best_splitpint_step.ImpurityBufferId,
best_splitpint_step.SplitpointBufferId,
best_splitpint_step.SplitpointCountsBufferId,
best_splitpint_step.ChildCountsEstimationBufferId,
best_splitpint_step.LeftEstimationYsBufferId,
best_splitpint_step.RightEstimationYsBufferId,
feature_params_step.FloatParamsBufferId,
feature_params_step.IntParamsBufferId,
matrix_feature_extractor_step.FeatureValuesBufferId,
feature_ordering,
matrix_feature_extractor_step)
should_split_criteria = create_should_split_criteria(**kwargs)
finalizer = regression.MeanVarianceEstimatorFinalizer_f32()
split_indices = splitpoints.SplitIndices_f32i32(sample_data_step.IndicesBufferId)
split_selector = splitpoints.SplitSelector_f32i32([split_buffers], should_split_criteria, finalizer, split_indices )
tree_learner = learn.BreadthFirstTreeLearner_f32i32(try_split_criteria, tree_steps_pipeline, node_steps_pipeline, split_selector, number_of_leaves)
forest_learner = learn.ParallelForestLearner(tree_learner, number_of_trees, dimension_of_y, number_of_jobs)
return forest_learner
def create_scaled_depth_delta_learner_32f(**kwargs):
ux = float(kwargs.get('ux'))
uy = float(kwargs.get('uy'))
vx = float(kwargs.get('vx'))
vy = float(kwargs.get('vy'))
number_of_trees = int(kwargs.get('number_of_trees', 10))
number_of_features = int(kwargs.get('number_of_features', 1))
feature_ordering = int(
kwargs.get('feature_ordering', pipeline.FEATURES_BY_DATAPOINTS))
number_of_jobs = int(kwargs.get('number_of_jobs', 1))
number_of_classes = int(kwargs['classes'].GetMax() + 1)
try_split_criteria = create_try_split_criteria(**kwargs)
if 'bootstrap' in kwargs and kwargs.get('bootstrap'):
sample_data_step = pipeline.BootstrapSamplesStep_i32f32i32(
buffers.PIXEL_INDICES)
else:
sample_data_step = pipeline.AllSamplesStep_i32f32i32(
buffers.PIXEL_INDICES)
number_of_features_buffer = buffers.as_vector_buffer(
np.array([number_of_features], dtype=np.int32))
set_number_features_step = pipeline.SetInt32VectorBufferStep(
number_of_features_buffer, pipeline.WHEN_NEW)
tree_steps_pipeline = pipeline.Pipeline(
[sample_data_step, set_number_features_step])
feature_params_step = image_features.PixelPairGaussianOffsetsStep_f32i32(
set_number_features_step.OutputBufferId, ux, uy, vx, vy)
depth_delta_feature = image_features.ScaledDepthDeltaFeature_f32i32(
feature_params_step.FloatParamsBufferId,
feature_params_step.IntParamsBufferId,
sample_data_step.IndicesBufferId, buffers.PIXEL_INDICES,
buffers.DEPTH_IMAGES, buffers.OFFSET_SCALES)
depth_delta_feature_extractor_step = image_features.ScaledDepthDeltaFeatureExtractorStep_f32i32(
depth_delta_feature, feature_ordering)
slice_classes_step = pipeline.SliceInt32VectorBufferStep_i32(
buffers.CLASS_LABELS, sample_data_step.IndicesBufferId)
slice_weights_step = pipeline.SliceFloat32VectorBufferStep_i32(
sample_data_step.WeightsBufferId, sample_data_step.IndicesBufferId)
class_infogain_walker = classification.ClassInfoGainWalker_f32i32(
slice_weights_step.SlicedBufferId, slice_classes_step.SlicedBufferId,
number_of_classes)
best_splitpint_step = classification.ClassInfoGainBestSplitpointsWalkingSortedStep_f32i32(
class_infogain_walker,
depth_delta_feature_extractor_step.FeatureValuesBufferId,
feature_ordering)
node_steps_pipeline = pipeline.Pipeline([
feature_params_step, depth_delta_feature_extractor_step,
slice_classes_step, slice_weights_step, best_splitpint_step
])
split_buffers = splitpoints.SplitSelectorBuffers(
best_splitpint_step.ImpurityBufferId,
best_splitpint_step.SplitpointBufferId,
best_splitpint_step.SplitpointCountsBufferId,
best_splitpint_step.ChildCountsBufferId,
best_splitpint_step.LeftYsBufferId,
best_splitpint_step.RightYsBufferId,
feature_params_step.FloatParamsBufferId,
feature_params_step.IntParamsBufferId,
depth_delta_feature_extractor_step.FeatureValuesBufferId,
feature_ordering, depth_delta_feature_extractor_step)
should_split_criteria = create_should_split_criteria(**kwargs)
finalizer = classification.ClassEstimatorFinalizer_f32()
split_indices = splitpoints.SplitIndices_f32i32(
sample_data_step.IndicesBufferId)
split_selector = splitpoints.SplitSelector_f32i32([split_buffers],
should_split_criteria,
finalizer, split_indices)
tree_learner = learn.DepthFirstTreeLearner_f32i32(try_split_criteria,
tree_steps_pipeline,
node_steps_pipeline,
split_selector)
forest_learner = learn.ParallelForestLearner(tree_learner, number_of_trees,
number_of_classes,
number_of_jobs)
return forest_learner
def create_biau2008_regression_scaled_depth_delta_learner_32f(**kwargs):
ux = float( kwargs.get('ux') )
uy = float( kwargs.get('uy') )
vx = float( kwargs.get('vx') )
vy = float( kwargs.get('vy') )
number_of_trees = int( kwargs.get('number_of_trees', 10) )
number_of_leaves = int( kwargs.get('number_of_leaves', kwargs['y'].GetM() / 5 + 1) )
number_of_split_retries = int( kwargs.get('number_of_split_retries', 10) )
number_of_features = 1
feature_ordering = int( kwargs.get('feature_ordering', pipeline.FEATURES_BY_DATAPOINTS) )
number_of_jobs = int( kwargs.get('number_of_jobs', 1) )
dimension_of_y = int( kwargs['y'].GetN() )
try_split_criteria = try_split.MinNodeSizeCriteria(2)
sample_data_step = pipeline.AllSamplesStep_i32f32i32(buffers.PIXEL_INDICES)
number_of_features_buffer = buffers.as_vector_buffer(np.array([number_of_features], dtype=np.int32))
set_number_features_step = pipeline.SetInt32VectorBufferStep(number_of_features_buffer, pipeline.WHEN_NEW)
tree_steps_pipeline = pipeline.Pipeline([sample_data_step, set_number_features_step])
feature_params_step = image_features.PixelPairGaussianOffsetsStep_f32i32(set_number_features_step.OutputBufferId, ux, uy, vx, vy )
depth_delta_feature = image_features.ScaledDepthDeltaFeature_f32i32(feature_params_step.FloatParamsBufferId,
feature_params_step.IntParamsBufferId,
sample_data_step.IndicesBufferId,
buffers.PIXEL_INDICES,
buffers.DEPTH_IMAGES,
buffers.OFFSET_SCALES)
depth_delta_feature_extractor_step = image_features.ScaledDepthDeltaFeatureExtractorStep_f32i32(depth_delta_feature, feature_ordering)
slice_ys_step = pipeline.SliceFloat32MatrixBufferStep_i32(buffers.YS, sample_data_step.IndicesBufferId)
slice_weights_step = pipeline.SliceFloat32VectorBufferStep_i32(sample_data_step.WeightsBufferId, sample_data_step.IndicesBufferId)
impurity_walker = regression.SumOfVarianceWalker_f32i32(slice_weights_step.SlicedBufferId,
slice_ys_step.SlicedBufferId,
dimension_of_y)
best_splitpint_step = regression.SumOfVarianceRandomGapSplitpointsStep_f32i32(impurity_walker,
depth_delta_feature_extractor_step.FeatureValuesBufferId,
feature_ordering)
node_steps_pipeline = pipeline.Pipeline([feature_params_step, depth_delta_feature_extractor_step,
slice_ys_step, slice_weights_step, best_splitpint_step])
split_buffers = splitpoints.SplitSelectorBuffers(best_splitpint_step.ImpurityBufferId,
best_splitpint_step.SplitpointBufferId,
best_splitpint_step.SplitpointCountsBufferId,
best_splitpint_step.ChildCountsBufferId,
best_splitpint_step.LeftYsBufferId,
best_splitpint_step.RightYsBufferId,
feature_params_step.FloatParamsBufferId,
feature_params_step.IntParamsBufferId,
depth_delta_feature_extractor_step.FeatureValuesBufferId,
feature_ordering,
depth_delta_feature_extractor_step)
should_split_criteria = min_child_size_criteria = should_split.MinChildSizeCriteria(1)
finalizer = regression.MeanVarianceEstimatorFinalizer_f32()
split_indices = splitpoints.SplitIndices_f32i32(sample_data_step.IndicesBufferId)
split_selector = splitpoints.SplitSelector_f32i32([split_buffers], should_split_criteria, finalizer, split_indices )
tree_learner = learn.Biau2008TreeLearner_f32i32(try_split_criteria, tree_steps_pipeline, node_steps_pipeline, split_selector, number_of_leaves, number_of_split_retries)
forest_learner = learn.ParallelForestLearner(tree_learner, number_of_trees, dimension_of_y, number_of_jobs)
return forest_learner
def tree_weights_for_all_trees(forest_size):
tree_weights=buffers.as_vector_buffer( np.ones(forest_size, dtype=np.float64) )
return tree_weights
def create_axis_aligned_matrix_walking_learner_32f(**kwargs):
number_of_trees = int(kwargs.get("number_of_trees", 10))
number_of_features = int(kwargs.get("number_of_features", np.sqrt(kwargs["x"].shape[1])))
feature_ordering = int(kwargs.get("feature_ordering", pipeline.FEATURES_BY_DATAPOINTS))
number_of_jobs = int(kwargs.get("number_of_jobs", 1))
number_of_classes = int(np.max(kwargs["classes"]) + 1)
try_split_criteria = create_try_split_criteria(**kwargs)
if "bootstrap" in kwargs and kwargs.get("bootstrap"):
sample_data_step = pipeline.BootstrapSamplesStep_f32f32i32(buffers.X_FLOAT_DATA)
else:
sample_data_step = pipeline.AllSamplesStep_f32f32i32(buffers.X_FLOAT_DATA)
number_of_features_buffer = buffers.as_vector_buffer(np.array([number_of_features], dtype=np.int32))
set_number_features_step = pipeline.SetInt32VectorBufferStep(number_of_features_buffer, pipeline.WHEN_NEW)
tree_steps_pipeline = pipeline.Pipeline([sample_data_step, set_number_features_step])
feature_params_step = matrix_features.AxisAlignedParamsStep_f32i32(
set_number_features_step.OutputBufferId, buffers.X_FLOAT_DATA
)
matrix_feature = matrix_features.LinearFloat32MatrixFeature_f32i32(
feature_params_step.FloatParamsBufferId,
feature_params_step.IntParamsBufferId,
sample_data_step.IndicesBufferId,
buffers.X_FLOAT_DATA,
)
matrix_feature_extractor_step = matrix_features.LinearFloat32MatrixFeatureExtractorStep_f32i32(
matrix_feature, feature_ordering
)
slice_classes_step = pipeline.SliceInt32VectorBufferStep_i32(buffers.CLASS_LABELS, sample_data_step.IndicesBufferId)
slice_weights_step = pipeline.SliceFloat32VectorBufferStep_i32(
sample_data_step.WeightsBufferId, sample_data_step.IndicesBufferId
)
class_infogain_walker = classification.ClassInfoGainWalker_f32i32(
slice_weights_step.SlicedBufferId, slice_classes_step.SlicedBufferId, number_of_classes
)
best_splitpint_step = classification.ClassInfoGainBestSplitpointsWalkingSortedStep_f32i32(
class_infogain_walker, matrix_feature_extractor_step.FeatureValuesBufferId, feature_ordering
)
node_steps_pipeline = pipeline.Pipeline(
[
feature_params_step,
matrix_feature_extractor_step,
slice_classes_step,
slice_weights_step,
best_splitpint_step,
]
)
split_buffers = splitpoints.SplitSelectorBuffers(
best_splitpint_step.ImpurityBufferId,
best_splitpint_step.SplitpointBufferId,
best_splitpint_step.SplitpointCountsBufferId,
best_splitpint_step.ChildCountsBufferId,
best_splitpint_step.LeftYsBufferId,
best_splitpint_step.RightYsBufferId,
feature_params_step.FloatParamsBufferId,
feature_params_step.IntParamsBufferId,
matrix_feature_extractor_step.FeatureValuesBufferId,
feature_ordering,
matrix_feature_extractor_step,
)
should_split_criteria = create_should_split_criteria(**kwargs)
finalizer = classification.ClassEstimatorFinalizer_f32()
split_indices = splitpoints.SplitIndices_f32i32(sample_data_step.IndicesBufferId)
split_selector = splitpoints.SplitSelector_f32i32([split_buffers], should_split_criteria, finalizer, split_indices)
if "tree_order" in kwargs and kwargs.get("tree_order") == "breadth_first":
tree_learner = learn.BreadthFirstTreeLearner_f32i32(
try_split_criteria, tree_steps_pipeline, node_steps_pipeline, split_selector
)
else:
tree_learner = learn.DepthFirstTreeLearner_f32i32(
try_split_criteria, tree_steps_pipeline, node_steps_pipeline, split_selector
)
forest_learner = learn.ParallelForestLearner(tree_learner, number_of_trees, number_of_classes, number_of_jobs)
return forest_learner
number_of_datapoints = pixel_indices_buffer.GetM()
offset_scales = np.array(np.random.uniform(0.8, 1.2, (number_of_datapoints, 2)), dtype=np.float32)
offset_scales_buffer = rftk.buffers.as_matrix_buffer(offset_scales)
# On the first pass through data learn for each sample counts
list_of_sample_counts = eval(args.list_of_sample_counts)
clipped_list_of_sample_counts = [min(s, pixel_labels_buffer.GetN()) for s in list_of_sample_counts]
clipped_list_of_sample_ranges = zip([0] + clipped_list_of_sample_counts[:-1], clipped_list_of_sample_counts)
print clipped_list_of_sample_ranges
pass_id = 0
for (start_index, end_index) in clipped_list_of_sample_ranges:
print start_index
print end_index
# Slice data
datapoint_indices = buffers.as_vector_buffer(np.array(np.arange(start_index, end_index), dtype=np.int32))
sliced_pixel_indices_buffer = pixel_indices_buffer.Slice(datapoint_indices)
sliced_offset_scales_buffer = offset_scales_buffer.Slice(datapoint_indices)
sliced_pixel_labels_buffer = pixel_labels_buffer.Slice(datapoint_indices)
# online_learner.Train(bufferCollection, buffers.Int32Vector(datapoint_indices))
predictor = forest_learner.fit(depth_images=depths_buffer,
pixel_indices=sliced_pixel_indices_buffer,
offset_scales=sliced_offset_scales_buffer,
classes=sliced_pixel_labels_buffer)
#pickle forest and data used for training
forest_pickle_filename = "%s/forest-%d-%d.pkl" % (online_run_folder, pass_id, end_index)
pickle.dump(predictor.get_forest(), gzip.open(forest_pickle_filename, 'wb'))
# Print forest stats
def create_online_axis_aligned_matrix_one_stream_learner_32f(**kwargs):
number_of_trees = int(kwargs.get("number_of_trees", 10))
number_of_features = int(kwargs.get("number_of_features", np.sqrt(kwargs["x"].shape[1])))
feature_ordering = int(kwargs.get("feature_ordering", pipeline.FEATURES_BY_DATAPOINTS))
number_of_splitpoints = int(kwargs.get("number_of_splitpoints", 1))
number_of_classes = int(np.max(kwargs["classes"]) + 1)
max_frontier_size = int(kwargs.get("max_frontier_size", 10000000))
impurity_update_period = int(kwargs.get("impurity_update_period", 1))
try_split_criteria = create_try_split_criteria(**kwargs)
if "bootstrap" in kwargs and kwargs.get("bootstrap"):
sample_data_step = pipeline.BootstrapSamplesStep_f32f32i32(buffers.X_FLOAT_DATA)
elif "poisson_sample" in kwargs:
poisson_sample_mean = float(kwargs.get("poisson_sample"))
sample_data_step = pipeline.PoissonSamplesStep_f32i32(buffers.X_FLOAT_DATA, poisson_sample_mean)
else:
sample_data_step = pipeline.AllSamplesStep_f32f32i32(buffers.X_FLOAT_DATA)
number_of_features_buffer = buffers.as_vector_buffer(np.array([number_of_features], dtype=np.int32))
set_number_features_step = pipeline.SetInt32VectorBufferStep(number_of_features_buffer, pipeline.WHEN_NEW)
tree_steps_pipeline = pipeline.Pipeline([sample_data_step, set_number_features_step])
feature_params_step = matrix_features.AxisAlignedParamsStep_f32i32(
set_number_features_step.OutputBufferId, buffers.X_FLOAT_DATA
)
matrix_feature = matrix_features.LinearFloat32MatrixFeature_f32i32(
feature_params_step.FloatParamsBufferId,
feature_params_step.IntParamsBufferId,
sample_data_step.IndicesBufferId,
buffers.X_FLOAT_DATA,
)
matrix_feature_extractor_step = matrix_features.LinearFloat32MatrixFeatureExtractorStep_f32i32(
matrix_feature, feature_ordering
)
slice_classes_step = pipeline.SliceInt32VectorBufferStep_i32(buffers.CLASS_LABELS, sample_data_step.IndicesBufferId)
slice_weights_step = pipeline.SliceFloat32VectorBufferStep_i32(
sample_data_step.WeightsBufferId, sample_data_step.IndicesBufferId
)
random_splitpoint_selection_step = splitpoints.RandomSplitpointsStep_f32i32(
matrix_feature_extractor_step.FeatureValuesBufferId, number_of_splitpoints, feature_ordering
)
class_stats_updater = classification.ClassStatsUpdater_f32i32(
slice_weights_step.SlicedBufferId, slice_classes_step.SlicedBufferId, number_of_classes
)
one_stream_split_stats_step = classification.ClassStatsUpdaterOneStreamStep_f32i32(
random_splitpoint_selection_step.SplitpointsBufferId,
random_splitpoint_selection_step.SplitpointsCountsBufferId,
matrix_feature_extractor_step.FeatureValuesBufferId,
feature_ordering,
class_stats_updater,
)
impurity_step = classification.ClassInfoGainSplitpointsImpurity_f32i32(
random_splitpoint_selection_step.SplitpointsCountsBufferId,
one_stream_split_stats_step.ChildCountsBufferId,
one_stream_split_stats_step.LeftStatsBufferId,
one_stream_split_stats_step.RightStatsBufferId,
)
init_node_steps_pipeline = pipeline.Pipeline([feature_params_step])
update_stats_node_steps_pipeline = pipeline.Pipeline(
[
matrix_feature_extractor_step,
slice_classes_step,
slice_weights_step,
random_splitpoint_selection_step,
one_stream_split_stats_step,
]
)
update_impurity_node_steps_pipeline = pipeline.Pipeline([impurity_step])
split_buffers = splitpoints.SplitSelectorBuffers(
impurity_step.ImpurityBufferId,
random_splitpoint_selection_step.SplitpointsBufferId,
random_splitpoint_selection_step.SplitpointsCountsBufferId,
one_stream_split_stats_step.ChildCountsBufferId,
one_stream_split_stats_step.LeftStatsBufferId,
one_stream_split_stats_step.RightStatsBufferId,
feature_params_step.FloatParamsBufferId,
feature_params_step.IntParamsBufferId,
matrix_feature_extractor_step.FeatureValuesBufferId,
feature_ordering,
matrix_feature_extractor_step,
)
should_split_criteria = create_should_split_criteria(**kwargs)
finalizer = classification.ClassEstimatorFinalizer_f32()
split_indices = splitpoints.SplitIndices_f32i32(sample_data_step.IndicesBufferId)
split_selector = splitpoints.WaitForBestSplitSelector_f32i32(
[split_buffers], should_split_criteria, finalizer, split_indices
)
matrix_feature_prediction = matrix_features.LinearFloat32MatrixFeature_f32i32(
sample_data_step.IndicesBufferId, buffers.X_FLOAT_DATA
)
estimator_params_updater = classification.ClassEstimatorUpdater_f32i32(
sample_data_step.WeightsBufferId, buffers.CLASS_LABELS, number_of_classes
)
forest_learner = learn.OnlineForestMatrixClassLearner_f32i32(
try_split_criteria,
tree_steps_pipeline,
init_node_steps_pipeline,
update_stats_node_steps_pipeline,
update_impurity_node_steps_pipeline,
impurity_update_period,
split_selector,
max_frontier_size,
number_of_trees,
5,
5,
number_of_classes,
sample_data_step.IndicesBufferId,
sample_data_step.WeightsBufferId,
matrix_feature_prediction,
estimator_params_updater,
)
return forest_learner
# On the first pass through data learn for each sample counts
list_of_sample_counts = eval(args.list_of_sample_counts)
clipped_list_of_sample_counts = [
min(s, pixel_labels_buffer.GetN()) for s in list_of_sample_counts
]
clipped_list_of_sample_ranges = zip([0] +
clipped_list_of_sample_counts[:-1],
clipped_list_of_sample_counts)
print clipped_list_of_sample_ranges
pass_id = 0
for (start_index, end_index) in clipped_list_of_sample_ranges:
print start_index
print end_index
# Slice data
datapoint_indices = buffers.as_vector_buffer(
np.array(np.arange(start_index, end_index), dtype=np.int32))
sliced_pixel_indices_buffer = pixel_indices_buffer.Slice(
datapoint_indices)
sliced_offset_scales_buffer = offset_scales_buffer.Slice(
datapoint_indices)
sliced_pixel_labels_buffer = pixel_labels_buffer.Slice(
datapoint_indices)
# online_learner.Train(bufferCollection, buffers.Int32Vector(datapoint_indices))
predictor = forest_learner.fit(
depth_images=depths_buffer,
pixel_indices=sliced_pixel_indices_buffer,
offset_scales=sliced_offset_scales_buffer,
classes=sliced_pixel_labels_buffer)
#pickle forest and data used for training
