def test_nfeat_no_feature_selection(self):
'''Test for n_feat when X.shape[1] < n_feat.'''
data = np.load('./test/testdata_basic.npz')
model_1d = fastl2lir.FastL2LiR()
model_2d = fastl2lir.FastL2LiR()
model_1d.fit(data['x_tr'], data['y_1d'], n_feat=500)
model_2d.fit(data['x_tr'], data['y_2d'], n_feat=500)
yp_1d = model_1d.predict(data['x_te'])
yp_2d = model_2d.predict(data['x_te'])
np.testing.assert_array_almost_equal(model_1d.W, data['w_1d'])
np.testing.assert_array_almost_equal(model_1d.b, data['b_1d'])
np.testing.assert_array_almost_equal(model_2d.W, data['w_2d'])
np.testing.assert_array_almost_equal(model_2d.b, data['b_2d'])
np.testing.assert_array_almost_equal(yp_1d, data['yp_1d'])
np.testing.assert_array_almost_equal(yp_2d, data['yp_2d'])
def test_nfeat(self):
'''Test for n_feat.'''
data = np.load('./test/testdata_nfeat.npz')
model_1d = fastl2lir.FastL2LiR()
model_2d = fastl2lir.FastL2LiR()
model_1d.fit(data['x_tr'], data['y_1d'], n_feat=20)
model_2d.fit(data['x_tr'], data['y_2d'], n_feat=20)
yp_1d = model_1d.predict(data['x_te'])
yp_2d = model_2d.predict(data['x_te'])
np.testing.assert_array_almost_equal(model_1d.W, data['w_1d'])
np.testing.assert_array_almost_equal(model_1d.b, data['b_1d'])
np.testing.assert_array_almost_equal(model_2d.W, data['w_2d'])
np.testing.assert_array_almost_equal(model_2d.b, data['b_2d'])
np.testing.assert_array_almost_equal(yp_1d, data['yp_1d'])
np.testing.assert_array_almost_equal(yp_2d, data['yp_2d'])
def test_alpha(self):
'''Test for alpha.'''
data = np.load('./test/testdata_alpha01.npz')
model_1d = fastl2lir.FastL2LiR()
model_2d = fastl2lir.FastL2LiR()
model_1d.fit(data['x_tr'], data['y_1d'], alpha=0.1)
model_2d.fit(data['x_tr'], data['y_2d'], alpha=0.1)
yp_1d = model_1d.predict(data['x_te'])
yp_2d = model_2d.predict(data['x_te'])
np.testing.assert_array_almost_equal(model_1d.W, data['w_1d'])
np.testing.assert_array_almost_equal(model_1d.b, data['b_1d'])
np.testing.assert_array_almost_equal(model_2d.W, data['w_2d'])
np.testing.assert_array_almost_equal(model_2d.b, data['b_2d'])
np.testing.assert_array_almost_equal(yp_1d, data['yp_1d'])
np.testing.assert_array_almost_equal(yp_2d, data['yp_2d'])
def test_chunk(self):
'''Test for chunk_size.'''
data = np.load('./test/testdata_chunk.npz')
model_2d = fastl2lir.FastL2LiR()
model_2d.fit(data['x_tr'], data['y_2d'], chunk_size=10)
yp_2d = model_2d.predict(data['x_te'])
np.testing.assert_array_almost_equal(model_2d.W, data['w_2d'])
np.testing.assert_array_almost_equal(model_2d.b, data['b_2d'])
np.testing.assert_array_almost_equal(yp_2d, data['yp_2d'])
def test_reshape(self):
'''Test for reshaping.'''
Y_shape = (200, 10, 10, 5)
X = np.random.rand(200, 100)
Y = np.random.rand(*Y_shape)
alpha = 1.0
n_feat = 50
model_test = fastl2lir.FastL2LiR()
model_test.fit(X, Y, alpha, n_feat)
pred_test = model_test.predict(X)
np.testing.assert_array_equal(model_test.W.shape, (100, ) + Y_shape[1:])
np.testing.assert_array_equal(model_test.b.shape, (1, ) + Y_shape[1:])
np.testing.assert_array_equal(pred_test.shape, Y_shape)
# Simple example #############################################################
# Preparation for simulation data
X = numpy.random.rand(600,
1000) # The number of samples x The number of voxels
Y = numpy.random.rand(
600, 500) # The number of samples x The number of CNN features
# Regularization parameter (coefficient for L2-norm)
alpha = 1.0
# The number of voxels after feature selection with correlation coefficients.
n_feat = 500
# Prepare a model object (like scikit-learn functions)
model = fastl2lir.FastL2LiR()
# Training
start_t = time.time()
model.fit(X, Y, alpha, n_feat)
print('Time for decoder training: ' + str(time.time() - start_t) + ' seconds')
# Prediction
Y_predicted = model.predict(X)
# Multi-dimensional target features ##########################################
X = numpy.random.rand(600, 1000)
# The target feature array is 3D.
Y = numpy.random.rand(600, 16, 16, 4)