search

caseykneale / ChemometricsTools.jl

A collection of tools for chemometrics and machine learning written in Julia.
Other
65 stars 12 forks source link

MultiPCA tensor product #11

Closed lnacquaroli closed 5 years ago

lnacquaroli commented 5 years ago

Hi there, sorry to bother again.

I have been snooping around the MultiPCA algorithm you used since before I found your library I was using drEEM from Bro et al. in Matlab.

I noticed the tensor product line https://github.com/caseykneale/ChemometricsTools.jl/blob/61c1f6724980329ed0fa27b03896e68c73a4f844/src/MultiWay.jl#L111

Tensors operations are quite developed in julia so far, in case you want to check them out: https://github.com/Jutho/TensorOperations.jl https://github.com/KristofferC/Tensors.jl https://github.com/ITensor/ITensors.jl

I confess I have not done any test on this, just to let you know for efficiency comparison in the future and to make use of the packages developed here.

caseykneale commented 5 years ago

Thanks for the heads up. When I first wrote this the exact tensor product I was performing there wasn't a direct operation I think. I tried doing it via the kronecker product onboard with julia and it exploded in memory for real data.

I forget the name of that specific product maybe there are other libraries out there (now) doing this sort of thing adequately.

caseykneale commented 5 years ago

Also let's leave this issue open - when I get more time I'd like to invest in the multiway algorithms. I have just been swamped.

lnacquaroli commented 5 years ago

Looking forward for the MultiwayPLS :)

caseykneale commented 5 years ago

I wrote that using Rasmus Bro's code (its on their website) years ago. I didn't like the way I structured it so when I do go for it this time around it'll be a lot more generic. I could probably spin up a version in an afternoon. The tricky part is getting the time! Maybe this weekend? No promises though.

caseykneale commented 5 years ago

Do you rely on multiway PLS right now? If so maybe the near future would be a good way to get some user feedback.

lnacquaroli commented 5 years ago

Not at the moment , I am not in a rush for it. I did use it a month ago...

On Tue, Aug 6, 2019 at 5:29 PM Casey Kneale notifications@github.com wrote:

Do you rely on multiway PLS right now?

— You are receiving this because you authored the thread.

Reply to this email directly, view it on GitHub https://github.com/caseykneale/ChemometricsTools.jl/issues/11?email_source=notifications&email_token=AGKCEEGPBQNMWGUYBDQU3TTQDHULXA5CNFSM4IIZ27H2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD3WQV2Q#issuecomment-518851306, or mute the thread https://github.com/notifications/unsubscribe-auth/AGKCEEHQYDUZESZKGVOPHQ3QDHULXANCNFSM4IIZ27HQ .

caseykneale commented 5 years ago

Garauntee there are some bugs but check master for an effort at MPLS

https://caseykneale.github.io/ChemometricsTools.jl/dev/man/MultiWay/

lnacquaroli commented 5 years ago

Sorry I did not have time to test it. I will follow up on a new issue if I ever have to. Thanks!