Enhance _cwt.py by introducing a configurable hop size parameter #804
Conversation
Generating a scalogram from the full output of the Continuous Wavelet Transform (CWT) entails high computational cost while providing limited performance gains in acoustic recognition models based on deep learning. Therefore, this update proposes reducing the output size during the intermediate processing stage—rather than after CWT generation—to improve computational efficiency of CWT. This pull request reflects the research findings presented in the following paper. Phan, D. T., Huynh, T. A., Pham, V. T., Tran, C. M., Mai, V. T., & Tran, N. Q. (2025). Optimal Scalogram for Computational Complexity Reduction in Acoustic Recognition Using Deep Learning. arXiv preprint arXiv:2505.13017.
There was a problem hiding this comment.
Thanks @phandangthoai. A few quick review comments inline. This change also needs tests, and it would be good to demonstrate what the gain of this is with a small benchmark.
|
Hello @rgommers, |
Update _cwt.py
|
Hello @rgommers,
|
There was a problem hiding this comment.
Thanks @phandangthoai. A few inline comments. In addtion, could you please:
- Share a benchmark script that shows the performance improvement
- Add a References section in the docstring to the paper in numpydoc format
pywt/_cwt.py
Outdated
| @@ -24,9 +24,9 @@ def next_fast_len(n): | |||
| return 2**ceil(np.log2(n)) | |||
|
|
|||
|
|
|||
| def cwt(data, scales, wavelet, sampling_period=1., method='conv', axis=-1): | |||
| def cwt(data, scales, wavelet, hop_size=1, sampling_period=1., method='conv', axis=-1): | |||
There was a problem hiding this comment.
Can you please move the new keyword to the end of the signature? It's not backwards-compatible to add a new keyword in the middle. That would break calls like cwt(data, scales, wavelet, 2.5) (2.5 was the sampling period).
Also, please add it as *, hop_size=1 (the * means keyword-only).
There was a problem hiding this comment.
The examples and tests then need to be updated for that change.
There was a problem hiding this comment.
- New key word hop_size is move to the end, and defined as keyword-only.
- The examples and tests are revised accordingly and passed.
Move hop_size to the end
Move hop_size to the end
Add References to the research paper.
Revise the notes
|
Hello @rgommers , Thank you for your valuable feedback. I have updated the pull request in accordance with your comments. The script for generating the optimal scalogram and the script for generating the standard scalogram have been included for reference. A References section citing the relevant paper has been added to the function docstring, following the numpydoc format. Please let me know if any further revisions are required. Thank you again for your time and guidance. Best regards, |
|
Hello @rgommers , I have addressed the trailing whitespace issue reported by the style check tool in my latest commit. However, it appears that the pre-commit style check did not re-trigger automatically. Could you kindly re-run it or let me know if there is anything further I need to do on my end? Best regards, |
|
This accumulated a merge conflict, could you please merge with I've triggered the CI to run. The added tests only use |
Fix trailing whitespaces error
|
The merge conflict is still unresolved, so I won't retrigger CI now. If you want to find a trivial typo or one word change somewhere that I can merge immediately in a separate PR, then CI jobs won't need manual approval from me after that. |
Test for hop_size greater than 1.
Resolve conflicts when merging to main
…-optimal-scalogram
…alogram 2 merge main into optimal scalogram
Test for hop size greater than 1
Test for hop size greater than 1
|
Hello @rgommers , I have completed the following updates to the PR:
Could you please review the changes and proceed with the merge into main? Best regards, |
Fix doctest error
|
Hello @rgommers, All checks have passed, and I am thrilled to be at the stage of having my first official PR ready for merge. Thank you for your time and support. Best regards, |
Generating a scalogram from the full output of the Continuous Wavelet Transform (CWT) entails high computational cost while providing limited performance gains in acoustic recognition models based on deep learning. Therefore, this update proposes reducing the output size during the intermediate processing stage—rather than after CWT generation—to improve computational efficiency of CWT. This pull request reflects the research findings presented in the following paper.
Phan, D. T., Huynh, T. A., Pham, V. T., Tran, C. M., Mai, V. T., & Tran, N. Q. (2025). Optimal Scalogram for Computational Complexity Reduction in Acoustic Recognition Using Deep Learning. arXiv preprint arXiv:2505.13017.