[ENH] Implemented Tracking differentiator-based multiview dilated characteristics (TD-MVDC) Classifier

[lucifer4073]

Reference Issues/PRs

Closes #2081.

What does this implement/fix? Explain your changes.

This PR implements the TDMVDC classifier.

Does your contribution introduce a new dependency? If yes, which one?

No

Any other comments?

Currently using the TSFreshRelevant feature extractor for feature extraction. I have tried implementing a similar module (for feature extraction) dedicated for tdmvdc itself (doesn't give great results).

Accuracy as per paper: 81.14 per cent Obtained accuracy: 77.71 per cent

PR checklist

For all contributions

• [X] I've added myself to the list of contributors. Alternatively, you can use the @all-contributors bot to do this for you after the PR has been merged.
• [X] The PR title starts with either [ENH], [MNT], [DOC], [BUG], [REF], [DEP] or [GOV] indicating whether the PR topic is related to enhancement, maintenance, documentation, bugs, refactoring, deprecation or governance.

For new estimators and functions

• [ ] I've added the estimator/function to the online API documentation.
• [ ] (OPTIONAL) I've added myself as a __maintainer__ at the top of relevant files and want to be contacted regarding its maintenance. Unmaintained files may be removed. This is for the full file, and you should not add yourself if you are just making minor changes or do not want to help maintain its contents.

For developers with write access

• [ ] (OPTIONAL) I've updated aeon's CODEOWNERS to receive notifications about future changes to these files.

[review-notebook-app[bot]]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[aeon-actions-bot[bot]]

Thank you for contributing to aeon

I have added the following labels to this PR based on the title: [ $\color{#FEF1BE}{\textsf{enhancement}}$ ]. I have added the following labels to this PR based on the changes made: [ $\color{#BCAE15}{\textsf{classification}}$, $\color{#45FD64}{\textsf{examples}}$, $\color{#41A8F6}{\textsf{transformations}}$ ]. Feel free to change these if they do not properly represent the PR.

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
• [ ] 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

[lucifer4073]

@CCHe64 this is the current implementation of the classifier. Other tasks such as testing and multithreading needs to be added. Will do that once the accuracy discrepancy is resolved.

[CCHe64]

@CCHe64 this is the current implementation of the classifier. Other tasks such as testing and multithreading needs to be added. Will do that once the accuracy discrepancy is resolved.

Is tsfreshrelevant used now? The original paper used 777 full feature versions of tsfresh. Is the accuracy difference because of this?

[lucifer4073]

Hi @MatthewMiddlehurst @TonyBagnall @CCHe64 ,

This is the current working implementation of TDMVDC. Further additions would be needed in test coverage and multithreading, which I would like to put up in another PR, once this is merged.

Let me know you reviews.

Thanks!

[lucifer4073]

It is unlikely we will merge this if it does not achieve similar results to the original.

Hi,

I have tested this of it on the load_arrow head dataset and it achieves the same accuracy as the paper.

[MatthewMiddlehurst]

Could you write a test or provide some example code and images showing that? Using more than one dataset would also be good.

[lucifer4073]

example code has been added for load_arrow_head. Will add some test cases for the same.

[CCHe64]

Thank you very much for your @TonyBagnall @MatthewMiddlehurst review and implementation of @lucifer4073 . I reviewed the code and performed classification experiments on the 85 UCR datasets. In the original paper, the average accuracy on 85 datasets is 84.13%. The Aeon implementation of TD-MVDC achieves an average accuracy of 84.07% on 85 UCR datasets. In addition, the number of win-lose datasets for both is very close. Therefore, I am happy with the Aeon version implementation. I recommend setting the default parameter of "feature_store_ratios" to [0.1, 0.2, 0.3, 0.4, 0.5] as in the original paper instead of None. I had no issues after making this change.

[MatthewMiddlehurst]

Great to hear it performs as expected. Thanks for running it. The code and testing still needs some improvements before this can be merged, however.

[lucifer4073]

Hi @CCHe64 @MatthewMiddlehurst ,

I have made the requested changes. As far the testing is concerned, I have only checked for a handful of datasets (accounting the execution time). Let me know if any improvements are needed.

[CCHe64]

Hi @CCHe64 @MatthewMiddlehurst ,

I have made the requested changes. As far the testing is concerned, I have only checked for a handful of datasets (accounting the execution time). Let me know if any improvements are needed.

I am happy with this implementation, thanks for your work.

[lucifer4073]

Hi @MatthewMiddlehurst ,

The latest changes have been updated

[lucifer4073]

Hi @MatthewMiddlehurst ,

I believe we can merge this now.