Fix quantal stdp test for numpy #681
Conversation
mschmidt87
force-pushed
the
fix_quantal_stdp_test_for_numpy
branch
from
1e3cf60
to
84b1252
Compare
There was a problem hiding this comment.
@mschmidt87 Please do not change the travis.yml file in this PR. I will comment on this in the issue, but it is a separate issue and requires careful consideration. I also added some small suggestions to the test.
| vm.shape = (n_trials, t_tot) | ||
| vm_reference.shape = (n_trials, t_tot) | ||
| vm.shape = (n_trials, int(t_tot)) | ||
| vm_reference.shape = (n_trials, int(t_tot)) |
There was a problem hiding this comment.
Using int(t_tot) here to dimension the result array is not a good idea, since you implicitly require that the voltmeter records at 1 ms intervals. A cleaner implementation would be (not 100% sure of spelling of assertions)
self.assert_equal(len(vm) % n_trials, 0)
n_steps = int(len(vm) / n_trials) # int to be absolutely Py3-safe
vm.shape = (n_trials, n_steps)
...| vm.shape = (n_trials, t_tot) | ||
| vm_reference.shape = (n_trials, t_tot) | ||
| vm.shape = (n_trials, int(t_tot)) | ||
| vm_reference.shape = (n_trials, int(t_tot)) | ||
|
|
||
| vm_mean = numpy.array([numpy.mean(vm[:, i]) | ||
| for i in range(int(t_tot))]) |
There was a problem hiding this comment.
I think you can avoid the list comprehension here, NumPy supports row- and column-wise means.
heplesser
added
ZC: PyNEST
Convert float variables to integers when using them as indices to numpy arrays and make reshaping of voltmeter more general
mschmidt87
force-pushed
the
fix_quantal_stdp_test_for_numpy
branch
from
84b1252
to
c4dfd24
Compare
|
Thank you for your feedback. I removed the commit touching .travis.yml from the PR and addressed your comments. I furthermore removed |
This pull request addresses #680 and contains the following changes:
t_tothad been implemented in this test at a different location earlier.