ValueError when attempting to run ExperimentBeijing with rank higher than 1 #2
Open
Description
I am attempting to execute make ExperimentBeijing with rank (
rPSMF/ExperimentBeijing/beijing_psmf.py
Line 102 in 93ac689
1 and get the following error:
Traceback (most recent call last):
File "/Users/ljoana/repos/rPSMF/ExperimentBeijing/beijing_psmf.py", line 171, in <module>
main()
File "/Users/ljoana/repos/rPSMF/ExperimentBeijing/beijing_psmf.py", line 151, in main
psmf.run(
File "/Users/ljoana/repos/rPSMF/ExperimentBeijing/beijing_psmf.py", line 71, in run
self.step(y_train, i, T)
File "/Users/ljoana/repos/rPSMF/pypsmf/psmf/psmf.py", line 77, in step
self.inner(i, k, y[k])
File "/Users/ljoana/repos/rPSMF/pypsmf/psmf/psmf.py", line 81, in inner
P_bar = self._predictive_covariance(i, k)
File "/Users/ljoana/repos/rPSMF/pypsmf/psmf/psmf.py", line 103, in _predictive_covariance
F = F.squeeze(axis=1).squeeze(axis=2)
ValueError: cannot select an axis to squeeze out which has size not equal to one
make: *** [ExperimentBeijing/output/periodic_bases.pdf] Error 1
I noticed that the shape of the nonlinearity output in FourierBasis
rPSMF/pypsmf/psmf/nonlinearities.py
Line 150 in 93ac689
(r, r) (based on the paper it should be (r, 1)), and that in the comments in _predictive_covariance() Lines 107 to 113 in 93ac689
The output of the Jacobian is (r, 1, r, 1), since the input, mu, to the nonlinearity is (r, 1) and it outputs an array of the same size..., but in reality the input is (r, r) and the Jacobian is (r, r, r, 1).Metadata
Assignees
Labels
No labels