Vincent Lostanlen

Results 194 comments of Vincent Lostanlen

Great summary! Thank you for writing it. A few additional remarks/questions 1. Quoting: > String data: string dtypes are technically [out of scope for the array api](https://data-apis.org/array-api/latest/purpose_and_scope.html#scope-includes-out-of-scope-non-goals). We use these...

Agree with @cyrusvahidi . Note that 2D does not introduce this extra dimension ```python batch_shape = input.shape[:-2] signal_shape = input.shape[-2:] input = input.reshape((-1,) + signal_shape) S = scattering2d(input, self.pad, self.unpad,...

Hello @chosw93, this sort of questions is best directed to a user group rather than an issue tracker. You are welcome to post them on the auditory@mcgill list http://www.auditory.org/

What is the status of this PR?

superseded by #925 (merged in August 2022)

@janden in the wake of https://github.com/kymatio/kymatio/pull/892, we no longer test torch 1.8 so i believe we could reopen this and merge it for v0.4

This passes tests, is backwards compatible, improves reported test coverage, fixes a known problem, and hasn't been reviewed in six months. Merging.

I'm reading this again and it's becoming difficult due to things like `average='global'` and phi/psi pairs in JTFS

i don't think it's worth doing. The sigma is very easy to read from `S.phi_f` if needed. But adding it to the meta is going to cause all sorts of...