Added Model for Time Series Classification by codeboy5 · Pull Request #253 · FluxML/FastAI.jl

codeboy5 · 2022-07-29T19:51:28Z

I have added the code for a basic RNN Model for the task of time series classification.

> data, blocks = load(datarecipes()["ecg5000"]);
> task = TSClassificationSingle(blocks, data);
> model = FastAI.taskmodel(task);
> traindl, validdl = taskdataloaders(data, task, 32);
> callbacks = [ToGPU(), Metrics(accuracy)];
> learner = Learner(model, tasklossfn(task); data=(traindl, validdl), optimizer=ADAM(), callbacks = callbacks);
> fitonecycle!(learner, 10, 0.033)

As I discussed with @darsnack, the idea is to add an encoding to do the reshaping to (features, batch, time) instead of doing it inside on the RNN Model. Working on that right now.

Have remove the type from StackedLSTM as it was redundant.

codeboy5 · 2022-07-30T04:22:07Z

I have also added a multivariate classification dataset and verified the training loop is working fine for it.

ToucheSir

Couple of drive-by comments on the WIP here

codeboy5 · 2022-08-11T16:46:32Z

Hey guys, @ToucheSir @darsnack @lorenzoh. I have updated the notebook. This PR essentially contains code to load regression datasets and the notebook for the classification task.
I will start adding the Inception Time Model in the next PR as we discussed during our weekly call.

ToucheSir

Looks good to go modulo a couple small things!

ToucheSir · 2022-09-01T04:34:05Z

+function blockmodel(inblock::TimeSeriesRow,
+                outblock::OneHotTensor{0}, 
+                backbone)
+    data   = rand(Float32, inblock.nfeatures, 32, inblock.obslength)


Should be using Flux.outputsize (I think we talked about this in a meeting). Changes required should be minimal :)

I tried doing that, but seems to give an error. Will try to debug it later.
https://pastebin.com/mB7aYvdD

I see, this is FluxML/Flux.jl#1755 (comment). Can you do the following then?

Use zeros or ones instead of rand.

Pass a batch size of 1 instead of 32 and seq length of 1 instead of inblock.obslength. This will save some time and you don't need it because you only care about the first output dimension.

reset! the backbone after you calculate output.

Yes, sounds great.

ToucheSir · 2022-09-01T04:35:41Z

 Random = "9a3f8284-a2c9-5f02-9a11-845980a1fd5c"
 Statistics = "10745b16-79ce-11e8-11f9-7d13ad32a3b2"
 UnicodePlots = "b8865327-cd53-5732-bb35-84acbb429228"
+Zygote = "e88e6eb3-aa80-5325-afca-941959d7151f"


Suggested change

Zygote = "e88e6eb3-aa80-5325-afca-941959d7151f"

I don't see Zygote used anywhere, so one less dependency doesn't hurt.

I tried doing that, but it throws an error

ERROR: LoadError: ArgumentError: Package FastTimeSeries does not have Zygote in its dependencies:`

This really could use a full stack trace, because AFAICT you're not using Zygote anywhere? Did you ] resolve after removing it from the project?

seems to work now, might have missed the resolve. thanks :)

lorenzoh · 2022-09-03T04:42:39Z

Great progress in my absence and cool to see the training working!

Are the outstanding comments being addressed in #256 or could this be merged individually beforehand?

codeboy5 · 2022-09-03T08:26:18Z

Great progress in my absence and cool to see the training working!

Are the outstanding comments being addressed in #256 or could this be merged individually beforehand?

I am resolving the outstanding comments in #256 , some of them have been resolved and a couple are left, which I will resolve by end of day. This can be merged then I believe.

codeboy5 · 2022-09-14T19:55:00Z

I think this PR is complete ?
Comments have been resolved in the next PR, let me know if something is missing.

ToucheSir · 2022-09-14T22:15:05Z

I wasn't aware #256 was also working on the StackedRNN model. It would be preferable to merge those changes here so that PR stays more focused on the InceptionTime. If not, we can just close this one and focus on that one (I'm interpreting your comment to mean that the changes in #256 completely subsume this PR).

codeboy5 · 2022-09-15T01:19:00Z

I wasn't aware #256 was also working on the StackedRNN model. It would be preferable to merge those changes here so that PR stays more focused on the InceptionTime. If not, we can just close this one and focus on that one (I'm interpreting your comment to mean that the changes in #256 completely subsume this PR).

PR #256 does contain the code changes in this PR, since I created that branch from this branch. Other than that, just some renaming that's done there, etc.
If we merge this, would those commits not be committed in that PR (since those commits are already committed to the master branch ) ?
Specifically Commits from added RNN Model to updated notebook.
If yes, we can merge this and then merge that one. Otherwise we can close this and merge that one.

ToucheSir · 2022-09-16T03:56:37Z

You'd likely have to rebase #256 after merging this PR if you've made changes to the same functions/types in that one. If that doesn't sound like a problem, we can merge this first.

codeboy5 · 2022-09-16T04:00:20Z

You'd likely have to rebase #256 after merging this PR if you've made changes to the same functions/types in that one. If that doesn't sound like a problem, we can merge this first.

Wouldn't be an issue. 👍🏻

ToucheSir

Merging as noted to follow up in #256.

codeboy5 added 2 commits July 30, 2022 01:13

added RNN Model

52a1932

added multivariate clf dataset

18cb1da

ToucheSir reviewed Jul 31, 2022

View reviewed changes

Comment thread FastTimeSeries/src/models/RNN.jl Outdated

Comment thread FastTimeSeries/src/models.jl Outdated

codeboy5 added 3 commits August 7, 2022 15:42

added blocks for regression task

ec6abe5

bug fixes (describetask)

f6834f6

updated notebook

54e8ebf

codeboy5 mentioned this pull request Aug 14, 2022

InceptionTime Model for Time Series #256

Merged

ToucheSir reviewed Sep 1, 2022

View reviewed changes

codeboy5 requested a review from ToucheSir September 14, 2022 19:54

ToucheSir approved these changes Sep 16, 2022

View reviewed changes

ToucheSir merged commit 3d70b50 into FluxML:master Sep 16, 2022

codeboy5 deleted the ts-models branch September 17, 2022 03:57

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Uh oh!

Added Model for Time Series Classification#253

Added Model for Time Series Classification#253
ToucheSir merged 5 commits into
FluxML:masterfrom
codeboy5:ts-models

codeboy5 commented Jul 29, 2022

codeboy5 commented Jul 30, 2022

ToucheSir left a comment

Uh oh!

Uh oh!

codeboy5 commented Aug 11, 2022

ToucheSir left a comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

ToucheSir Sep 1, 2022

codeboy5 Sep 4, 2022

ToucheSir Sep 4, 2022

codeboy5 Sep 4, 2022

ToucheSir Sep 1, 2022

codeboy5 Sep 4, 2022

ToucheSir Sep 4, 2022

codeboy5 Sep 4, 2022

Uh oh!

lorenzoh commented Sep 3, 2022

codeboy5 commented Sep 3, 2022

codeboy5 commented Sep 14, 2022

ToucheSir commented Sep 14, 2022

codeboy5 commented Sep 15, 2022

ToucheSir commented Sep 16, 2022

codeboy5 commented Sep 16, 2022

ToucheSir left a comment

Labels

3 participants

Uh oh!

Uh oh!

Conversation

codeboy5 commented Jul 29, 2022

codeboy5 commented Jul 30, 2022

ToucheSir left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

codeboy5 commented Aug 11, 2022

ToucheSir left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Uh oh!

lorenzoh commented Sep 3, 2022

codeboy5 commented Sep 3, 2022

codeboy5 commented Sep 14, 2022

ToucheSir commented Sep 14, 2022

codeboy5 commented Sep 15, 2022

ToucheSir commented Sep 16, 2022

codeboy5 commented Sep 16, 2022

ToucheSir left a comment

Choose a reason for hiding this comment

Labels

3 participants