def test_running_of_different_activation_modes(self):
#just tests that things run, not a test for values
for mode in [blobs.NonlinearMxtsMode.vals]:
out_layer = blobs.ReLU(
nonlinear_mxts_mode=blobs.NonlinearMxtsMode.DeepLIFT)
fprop_results, bprop_results_each_task =\
self.set_up_prediction_func_and_deeplift_func(out_layer)
self.assertListEqual(fprop_results, [[9.0, 0.0], [19.0, 0.0]])
def test_relu_after_dense(self):
input_layer = blobs.Input(num_dims=None, shape=(None, 4))
dense_layer = blobs.Dense(W=np.random.random((2, 4)),
b=np.random.random((2, )),
dense_mxts_mode=DenseMxtsMode.Linear)
dense_layer.set_inputs(input_layer)
relu_after_dense = blobs.ReLU(
nonlinear_mxts_mode=NonlinearMxtsMode.DeepLIFT_GenomicsDefault)
relu_after_dense.set_inputs(dense_layer)
relu_after_dense.build_fwd_pass_vars()
self.assertEqual(relu_after_dense.nonlinear_mxts_mode,
NonlinearMxtsMode.RevealCancel)
def test_relu_gradient(self):
out_layer = blobs.ReLU(
nonlinear_mxts_mode=blobs.NonlinearMxtsMode.Gradient)
fprop_results, bprop_results_each_task =\
self.set_up_prediction_func_and_deeplift_func(out_layer)
np.testing.assert_almost_equal(
np.array(bprop_results_each_task[0]),
np.array([np.array(self.w1), np.array(self.w1)]))
np.testing.assert_almost_equal(
np.array(bprop_results_each_task[1]),
np.array([np.zeros_like(self.w2),
np.zeros_like(self.w2)]))
def test_relu_intgrad(self):
out_layer = blobs.ReLU(
nonlinear_mxts_mode=blobs.NonlinearMxtsMode.Gradient)
out_layer.set_inputs(self.dense_layer)
out_layer.build_fwd_pass_vars()
self.input_layer.reset_mxts_updated()
out_layer.set_scoring_mode(blobs.ScoringMode.OneAndZeros)
out_layer.set_active()
self.input_layer.update_mxts()
fprop_func = compile_func([self.input_layer.get_activation_vars()],
out_layer.get_activation_vars())
fprop_results = [list(x) for x in fprop_func(self.inp)]
grad_func_temp = compile_func([
self.input_layer.get_activation_vars(),
self.input_layer.get_reference_vars()
], self.input_layer.get_mxts())
grad_func = (
lambda input_data_list, input_references_list, **kwargs:
#index [0] below is because we are retrieving the
#first mode, to be passed into grad_func_temp
grad_func_temp(input_data_list[0], input_references_list[0]))
integrated_grads_func = (
deeplift.util.get_integrated_gradients_function(
gradient_computation_function=grad_func, num_intervals=10))
bprop_results_each_task = []
for task_idx in range(len(fprop_results[0])):
out_layer.update_task_index(task_index=task_idx) #set task
bprop_results_task = [
list(x) for x in integrated_grads_func(
task_idx=None, #task setting handled manually
#in line above
input_data_list=np.array([self.inp]),
input_references_list=np.array(
[0.1 * np.ones_like(self.inp)]),
#batch_size and progress_update
#are ignored by grad_func
batch_size=20,
progress_update=10)
]
bprop_results_each_task.append(bprop_results_task)
out_layer.set_inactive()
print(bprop_results_each_task)
np.testing.assert_almost_equal(
np.array(bprop_results_each_task[0]),
np.array([-2.0 / 5.0 * np.array([1.0, 4.0])]))
def test_relu_after_conv2d(self):
input_layer = blobs.Input(num_dims=None, shape=(None, 2, 2, 2))
conv_layer = blobs.Conv2D(W=np.random.random((2, 2, 2, 2)),
b=np.random.random((2, )),
conv_mxts_mode=ConvMxtsMode.Linear,
strides=(1, 1),
border_mode=PaddingMode.valid,
channels_come_last=True)
conv_layer.set_inputs(input_layer)
relu_after_conv = blobs.ReLU(
nonlinear_mxts_mode=NonlinearMxtsMode.DeepLIFT_GenomicsDefault)
relu_after_conv.set_inputs(conv_layer)
relu_after_conv.build_fwd_pass_vars()
self.assertEqual(relu_after_conv.nonlinear_mxts_mode,
NonlinearMxtsMode.Rescale)
def test_relu_after_dense_batchnorm(self):
input_layer = blobs.Input(num_dims=None, shape=(None, 4))
dense_layer = blobs.Dense(W=np.random.random((4, 2)),
b=np.random.random((2, )),
dense_mxts_mode=DenseMxtsMode.Linear)
dense_layer.set_inputs(input_layer)
batch_norm = blobs.BatchNormalization(gamma=np.array([1.0, 1.0]),
beta=np.array([-0.5, 0.5]),
axis=-1,
mean=np.array([-0.5, 0.5]),
std=np.array([1.0, 1.0]),
epsilon=0.001)
batch_norm.set_inputs(dense_layer)
relu_after_bn = blobs.ReLU(
nonlinear_mxts_mode=NonlinearMxtsMode.DeepLIFT_GenomicsDefault)
relu_after_bn.set_inputs(batch_norm)
relu_after_bn.build_fwd_pass_vars()
self.assertEqual(relu_after_bn.nonlinear_mxts_mode,
NonlinearMxtsMode.RevealCancel)
def test_relu_deeplift(self):
out_layer = blobs.ReLU(
nonlinear_mxts_mode=blobs.NonlinearMxtsMode.DeepLIFT)
fprop_results, bprop_results_each_task =\
self.set_up_prediction_func_and_deeplift_func(out_layer)
self.assertListEqual(fprop_results, [[9.0, 0.0], [19.0, 0.0]])
#post-activation under default would be [0.0, 1.0]
#post-activation diff from default = [9.0, -1.0], [19.0, -1.0]
#pre-activation under default would be [-1.0, 1.0]
#pre-activation diff-from-default is [10.0, -10.0], [20.0, -20.0]
#scale-factors: [[9.0/10.0, -1.0/-10.0], [19.0/20.0, -1.0/-20.0]]
print(bprop_results_each_task)
np.testing.assert_almost_equal(
np.array(bprop_results_each_task[0]),
np.array([(9.0 / 10.0) * np.array(self.w1),
(19.0 / 20.0) * np.array(self.w1)]))
np.testing.assert_almost_equal(
np.array(bprop_results_each_task[1]),
np.array([(-1.0 / -10.0) * np.array(self.w2),
(-1.0 / -20.0) * np.array(self.w2)]))
def test_relu_after_conv1d_batchnorm(self):
input_layer = blobs.Input(num_dims=None, shape=(None, 2, 2))
conv_layer = blobs.Conv1D(W=np.random.random((2, 2, 2)),
b=np.random.random((2, )),
conv_mxts_mode=ConvMxtsMode.Linear,
stride=1,
border_mode=PaddingMode.valid,
channels_come_last=True)
conv_layer.set_inputs(input_layer)
batch_norm = blobs.BatchNormalization(gamma=np.array([1.0, 1.0]),
beta=np.array([-0.5, 0.5]),
axis=-1,
mean=np.array([-0.5, 0.5]),
std=np.array([1.0, 1.0]),
epsilon=0.001)
batch_norm.set_inputs(conv_layer)
relu_after_bn = blobs.ReLU(
nonlinear_mxts_mode=NonlinearMxtsMode.DeepLIFT_GenomicsDefault)
relu_after_bn.set_inputs(batch_norm)
relu_after_bn.build_fwd_pass_vars()
self.assertEqual(relu_after_bn.nonlinear_mxts_mode,
NonlinearMxtsMode.Rescale)
def test_relu_revealcancel(self):
out_layer = blobs.ReLU(
nonlinear_mxts_mode=blobs.NonlinearMxtsMode.RevealCancel)
fprop_results, bprop_results_each_task =\
self.set_up_prediction_func_and_deeplift_func(out_layer)
self.assertListEqual(fprop_results, [[3.0]])
#-1.0, 2.0, -3.0, 4.0: -4.0 and 6.0
#-1.0, 2.0, -3.0, 4.0: -4.0 and 6.0
#post-activation under reference: 1.0
#pre-activation diff from default = -: -4.0, +: 6.0
#post-activation diff-from-default = -: 0.5(-1.0 + -4.0) = -2.5
# +: 0.5(3.0 + 6.0) = 4.5
#multiplier: 2.5/4.0 and 4.5/6.0
neg_mult = 2.5 / 4.0
pos_mult = 4.5 / 6.0
print(bprop_results_each_task)
np.testing.assert_almost_equal(
np.array(bprop_results_each_task[0]),
np.array([
np.array([neg_mult, pos_mult, neg_mult, pos_mult]) *
np.array(self.w1)
]))
def relu_conversion(layer, name, verbose, nonlinear_mxts_mode):
return [
blobs.ReLU(name=name,
verbose=verbose,
nonlinear_mxts_mode=nonlinear_mxts_mode)
]
def relu_conversion(layer, name, mxts_mode):
return [blobs.ReLU(name=name, mxts_mode=mxts_mode)]
def test_relu_after_other_layer(self):
input_layer = blobs.Input(num_dims=None, shape=(None, 4))
relu_layer = blobs.ReLU(
nonlinear_mxts_mode=NonlinearMxtsMode.DeepLIFT_GenomicsDefault)
relu_layer.set_inputs(input_layer)
relu_layer.build_fwd_pass_vars()
def test_running_of_different_activation_modes(self):
#just tests that things run, not a test for values
for mode in blobs.NonlinearMxtsMode.vals:
out_layer = blobs.ReLU(nonlinear_mxts_mode=mode)
fprop_results, bprop_results_each_task =\
self.set_up_prediction_func_and_deeplift_func(out_layer)
