def get_weights_topo(self):
if not isinstance(self.input_space, Conv2DSpace):
raise NotImplementedError()
desired = self.W.get_value().T
ipt = self.desired_space.format_as(desired, self.input_space)
rval = Conv2DSpace.convert_numpy(ipt, self.input_space.axes, ('b', 0, 1, 'c'))
return rval
def get_weights_topo(self):
if not isinstance(self.input_space, Conv2DSpace):
raise NotImplementedError()
desired = self.W.get_value().T
ipt = self.desired_space.format_as(desired, self.input_space)
rval = Conv2DSpace.convert_numpy(ipt, self.input_space.axes, ('b', 0, 1, 'c'))
return rval
def convert_to_axes(reference_result, axes):
# assuming we have b c 0 1 (2) for reference
if (reference_result.ndim == 4):
return Conv2DSpace.convert_numpy(reference_result, ('b', 'c', 0, 1),
axes)
elif (reference_result.ndim == 5):
return Conv3DSpace.convert_numpy(reference_result, ('b', 'c', 0, 1, 2),
axes)
else:
raise ValueError(("Expect result to have 4 or 5 dims, "
"has {:d} dims".format(reference_result.ndim)))
def convert_to_axes(reference_result, axes):
# assuming we have b c 0 1 (2) for reference
if (reference_result.ndim == 4):
return Conv2DSpace.convert_numpy(reference_result,
('b', 'c', 0, 1), axes)
elif (reference_result.ndim == 5):
return Conv3DSpace.convert_numpy(reference_result,
('b', 'c', 0, 1, 2), axes)
else:
raise ValueError(("Expect result to have 4 or 5 dims, "
"has {:d} dims".format(reference_result.ndim)))
design_X = X
design_X_sequence = X_sequence
zeroed = (1.-drop_mask) * X
zeroed = dataset.get_topological_view(dataset.mapback_for_viewer(dataset.get_design_matrix(zeroed)))
M = dataset.get_topological_view(dataset.mapback_for_viewer(design_X))
print (M.min(), M.max())
M_sequence = [ dataset.get_topological_view(dataset.mapback_for_viewer(mat)) for mat in design_X_sequence ]
X = dataset.adjust_to_be_viewed_with(Xt,Xt,per_example=True)
if X_sequence[0].ndim == 2:
X_sequence = [ dataset.get_topological_view(mat) for mat in X_sequence ]
else:
model_axes = model.get_input_space().axes
display_axes = ('b', 0, 1, 'c')
X = Conv2DSpace.convert_numpy(X, model_axes, display_axes)
Xt = Conv2DSpace.convert_numpy(Xt, model_axes, display_axes)
drop_mask = Conv2DSpace.convert_numpy(drop_mask, model_axes, display_axes)
X_sequence = [Conv2DSpace.convert_numpy(elem, model_axes, display_axes) for elem in X_sequence]
X_sequence = [ dataset.adjust_to_be_viewed_with(mat,Xt,per_example=True) for mat in X_sequence ]
if cost.supervised:
rows += m
drop_mask_Y = outputs[end_X_outputs]
Y_sequence = outputs[end_X_outputs+1:]
pv = PatchViewer( (rows, cols), (X.shape[1], X.shape[2]), is_color = True,
pad = (8,8) )
dataset.mapback_for_viewer(dataset.get_design_matrix(zeroed)))
M = dataset.get_topological_view(dataset.mapback_for_viewer(design_X))
print(M.min(), M.max())
M_sequence = [
dataset.get_topological_view(dataset.mapback_for_viewer(mat))
for mat in design_X_sequence
]
X = dataset.adjust_to_be_viewed_with(Xt, Xt, per_example=True)
if X_sequence[0].ndim == 2:
X_sequence = [dataset.get_topological_view(mat) for mat in X_sequence]
else:
model_axes = model.get_input_space().axes
display_axes = ('b', 0, 1, 'c')
X = Conv2DSpace.convert_numpy(X, model_axes, display_axes)
Xt = Conv2DSpace.convert_numpy(Xt, model_axes, display_axes)
drop_mask = Conv2DSpace.convert_numpy(drop_mask, model_axes,
display_axes)
X_sequence = [
Conv2DSpace.convert_numpy(elem, model_axes, display_axes)
for elem in X_sequence
]
X_sequence = [
dataset.adjust_to_be_viewed_with(mat, Xt, per_example=True)
for mat in X_sequence
]
if cost.supervised:
rows += m
