[MNT] Kill TapNet with fire

hadifawaz1999 commented 1 week ago

our implementation of tapnet is wrong, i prefer not re-work and simply restart from scratch at some point, finding errors line by line will consume more energy and time than starting from scratch as i recall lots of it were wrongly coded. Prefer purging it before 1.0

Issues i found in it:

implementation of classifier and regressor not inline with all others, needs a huge refactor to be able to save/load, use always best model, random state usage
usage of lambda layers which is no longer allowed, such as for Encoder we had to change it
vary much bits of code that are not used just declared like : "conv_1_models" which is a declared keras sequential model but not used, "euclidean_dist" too
the convolution operations are not applied as the paper is saying it should, the original way is doing RDP then one convolution block on each Group, concatenate them and share the rest of the network, the current implementation does the whole network on each Group independently. The usage of keras-self attention should be stopped as many new ways in keras exist now to do it, as keras self attention is no longer reliable repository archive
RDP implementation is not efficient with tf.gather, making custom layers with no trainable parameters would be more appropriate

For more information on TapNet, the original paper

aeon-actions-bot[bot] commented 1 week ago

Thank you for contributing to `aeon`

I have added the following labels to this PR based on the title: [ $\color{#EC843A}{\textsf{maintenance}}$ ]. This PR changes too many different packages (>3) for automatic addition of labels, please manually add package labels if relevant.

The Checks tab will show the status of our automated tests. You can click on individual test runs in the tab or "Details" in the panel below to see more information if there is a failure.

If our pre-commit code quality check fails, any trivial fixes will automatically be pushed to your PR unless it is a draft.

Don't hesitate to ask questions on the aeon Slack channel if you have any.

PR CI actions

These checkboxes will add labels to enable/disable CI functionality for this PR. This may not take effect immediately, and a new commit may be required to run the new configuration.

[ ] Run pre-commit checks for all files
[ ] Run mypy typecheck tests
[ ] Run all pytest tests and configurations
[x] Run all notebook example tests
[ ] Run numba-disabled codecov tests
[ ] Stop automatic pre-commit fixes (always disabled for drafts)
[ ] Disable numba cache loading
[ ] Push an empty commit to re-run CI checks

TonyBagnall commented 1 week ago

we have discussed this off and on for a while so it makes sense to do this now before version 1.0

TonyBagnall commented 1 week ago

removing the soft dependency on keras-self attention is a big plus too

hadifawaz1999 commented 1 week ago

TODO: remove tapenet from notebook results loaders

review-notebook-app[bot] commented 1 week ago

Check out this pull request on

See visual diffs & provide feedback on Jupyter Notebooks.

Powered by ReviewNB

aeon-toolkit / aeon