fast-MPN-COV icon indicating copy to clipboard operation
fast-MPN-COV copied to clipboard

Published 20 hours ago • verified publisher icon jiangtaoxie

1D version of Fast-MPN-Cov

Open kayuksel opened this issue 6 years ago • 5 comments

Can you please also release a 1D version that we can use for time series rather than images?

kayuksel avatar Feb 15 '19 19:02 kayuksel

@kayuksel Before computing the correlation between channels, we first flatten the 2D feature map (HxW) into 1D vector (HW). I think you can skip this operation and directly compute covariance matrix of 1D time series signal. btw, it's a good suggestion for us to extend our method compatible with 1D data, I'll commit this improvement soon.

jiangtaoxie avatar Feb 16 '19 14:02 jiangtaoxie

@jiangtaoxie GAP is frequently used also in SoA architectures for 1D data, hence I want to try iSQRT-COV.

kayuksel avatar Feb 16 '19 18:02 kayuksel

@jiangtoxie Please check the following issue. It seems possible to get this into PyTorch, maybe if you create a pull request.

https://github.com/pytorch/pytorch/issues/17169

kayuksel avatar Mar 07 '19 06:03 kayuksel

@kayuksel That's great, Thanks a lot! I'll create a PR for that.

jiangtaoxie avatar Mar 07 '19 06:03 jiangtaoxie

@jiangtaoxie Do you have news about the 1D version yet?

akaniklaus avatar Jun 25 '19 15:06 akaniklaus