Unveiling The Mask of Position-Information Pattern Through the Mist of Image Features

Chieh Hubert Lin, Hung-Yu Tseng, Hsin-Ying Lee, Maneesh Kumar Singh, Ming-Hsuan Yang
Proceedings of the 40th International Conference on Machine Learning, PMLR 202:21204-21222, 2023.

Abstract

Recent studies have shown that paddings in convolutional neural networks encode absolute position information which can negatively affect the model performance for certain tasks. However, existing metrics for quantifying the strength of positional information remain unreliable and frequently lead to erroneous results. To address this issue, we propose novel metrics for measuring and visualizing the encoded positional information. We formally define the encoded information as Position-information Pattern from Padding (PPP) and conduct a series of experiments to study its properties as well as its formation. The proposed metrics measure the presence of positional information more reliably than the existing metrics based on PosENet and tests in F-Conv. We also demonstrate that for any extant (and proposed) padding schemes, PPP is primarily a learning artifact and is less dependent on the characteristics of the underlying padding schemes.

Cite this Paper

BibTeX
@InProceedings{pmlr-v202-lin23k, title = {Unveiling The Mask of Position-Information Pattern Through the Mist of Image Features}, author = {Lin, Chieh Hubert and Tseng, Hung-Yu and Lee, Hsin-Ying and Singh, Maneesh Kumar and Yang, Ming-Hsuan}, booktitle = {Proceedings of the 40th International Conference on Machine Learning}, pages = {21204--21222}, year = {2023}, editor = {Krause, Andreas and Brunskill, Emma and Cho, Kyunghyun and Engelhardt, Barbara and Sabato, Sivan and Scarlett, Jonathan}, volume = {202}, series = {Proceedings of Machine Learning Research}, month = {23--29 Jul}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v202/lin23k/lin23k.pdf}, url = {https://proceedings.mlr.press/v202/lin23k.html}, abstract = {Recent studies have shown that paddings in convolutional neural networks encode absolute position information which can negatively affect the model performance for certain tasks. However, existing metrics for quantifying the strength of positional information remain unreliable and frequently lead to erroneous results. To address this issue, we propose novel metrics for measuring and visualizing the encoded positional information. We formally define the encoded information as Position-information Pattern from Padding (PPP) and conduct a series of experiments to study its properties as well as its formation. The proposed metrics measure the presence of positional information more reliably than the existing metrics based on PosENet and tests in F-Conv. We also demonstrate that for any extant (and proposed) padding schemes, PPP is primarily a learning artifact and is less dependent on the characteristics of the underlying padding schemes.} }
Endnote
%0 Conference Paper %T Unveiling The Mask of Position-Information Pattern Through the Mist of Image Features %A Chieh Hubert Lin %A Hung-Yu Tseng %A Hsin-Ying Lee %A Maneesh Kumar Singh %A Ming-Hsuan Yang %B Proceedings of the 40th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2023 %E Andreas Krause %E Emma Brunskill %E Kyunghyun Cho %E Barbara Engelhardt %E Sivan Sabato %E Jonathan Scarlett %F pmlr-v202-lin23k %I PMLR %P 21204--21222 %U https://proceedings.mlr.press/v202/lin23k.html %V 202 %X Recent studies have shown that paddings in convolutional neural networks encode absolute position information which can negatively affect the model performance for certain tasks. However, existing metrics for quantifying the strength of positional information remain unreliable and frequently lead to erroneous results. To address this issue, we propose novel metrics for measuring and visualizing the encoded positional information. We formally define the encoded information as Position-information Pattern from Padding (PPP) and conduct a series of experiments to study its properties as well as its formation. The proposed metrics measure the presence of positional information more reliably than the existing metrics based on PosENet and tests in F-Conv. We also demonstrate that for any extant (and proposed) padding schemes, PPP is primarily a learning artifact and is less dependent on the characteristics of the underlying padding schemes.
APA
Lin, C.H., Tseng, H., Lee, H., Singh, M.K. & Yang, M.. (2023). Unveiling The Mask of Position-Information Pattern Through the Mist of Image Features. Proceedings of the 40th International Conference on Machine Learning, in Proceedings of Machine Learning Research 202:21204-21222 Available from https://proceedings.mlr.press/v202/lin23k.html.

Related Material