A deep convolutional neural network that is invariant to time rescaling

Brandon G Jacques, Zoran Tiganj, Aakash Sarkar, Marc Howard, Per Sederberg
Proceedings of the 39th International Conference on Machine Learning, PMLR 162:9729-9738, 2022.

Abstract

Human learners can readily understand speech, or a melody, when it is presented slower or faster than usual. This paper presents a deep CNN (SITHCon) that uses a logarithmically compressed temporal representation at each level. Because rescaling the time of the input results in a translation of $\log$ time, and because the output of the convolution is invariant to translations, this network can generalize to out-of-sample data that are temporal rescalings of a learned pattern. We compare the performance of SITHCon to a Temporal Convolution Network (TCN) on classification and regression problems with both univariate and multivariate time series. We find that SITHCon, unlike TCN, generalizes robustly over rescalings of about an order of magnitude. Moreover, we show that the network can generalize over exponentially large scales without retraining the weights simply by extending the range of the logarithmically-compressed temporal memory.

Cite this Paper

BibTeX
@InProceedings{pmlr-v162-jacques22a, title = {A deep convolutional neural network that is invariant to time rescaling}, author = {Jacques, Brandon G and Tiganj, Zoran and Sarkar, Aakash and Howard, Marc and Sederberg, Per}, booktitle = {Proceedings of the 39th International Conference on Machine Learning}, pages = {9729--9738}, year = {2022}, editor = {Chaudhuri, Kamalika and Jegelka, Stefanie and Song, Le and Szepesvari, Csaba and Niu, Gang and Sabato, Sivan}, volume = {162}, series = {Proceedings of Machine Learning Research}, month = {17--23 Jul}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v162/jacques22a/jacques22a.pdf}, url = {https://proceedings.mlr.press/v162/jacques22a.html}, abstract = {Human learners can readily understand speech, or a melody, when it is presented slower or faster than usual. This paper presents a deep CNN (SITHCon) that uses a logarithmically compressed temporal representation at each level. Because rescaling the time of the input results in a translation of $\log$ time, and because the output of the convolution is invariant to translations, this network can generalize to out-of-sample data that are temporal rescalings of a learned pattern. We compare the performance of SITHCon to a Temporal Convolution Network (TCN) on classification and regression problems with both univariate and multivariate time series. We find that SITHCon, unlike TCN, generalizes robustly over rescalings of about an order of magnitude. Moreover, we show that the network can generalize over exponentially large scales without retraining the weights simply by extending the range of the logarithmically-compressed temporal memory.} }
Endnote
%0 Conference Paper %T A deep convolutional neural network that is invariant to time rescaling %A Brandon G Jacques %A Zoran Tiganj %A Aakash Sarkar %A Marc Howard %A Per Sederberg %B Proceedings of the 39th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2022 %E Kamalika Chaudhuri %E Stefanie Jegelka %E Le Song %E Csaba Szepesvari %E Gang Niu %E Sivan Sabato %F pmlr-v162-jacques22a %I PMLR %P 9729--9738 %U https://proceedings.mlr.press/v162/jacques22a.html %V 162 %X Human learners can readily understand speech, or a melody, when it is presented slower or faster than usual. This paper presents a deep CNN (SITHCon) that uses a logarithmically compressed temporal representation at each level. Because rescaling the time of the input results in a translation of $\log$ time, and because the output of the convolution is invariant to translations, this network can generalize to out-of-sample data that are temporal rescalings of a learned pattern. We compare the performance of SITHCon to a Temporal Convolution Network (TCN) on classification and regression problems with both univariate and multivariate time series. We find that SITHCon, unlike TCN, generalizes robustly over rescalings of about an order of magnitude. Moreover, we show that the network can generalize over exponentially large scales without retraining the weights simply by extending the range of the logarithmically-compressed temporal memory.
APA
Jacques, B.G., Tiganj, Z., Sarkar, A., Howard, M. & Sederberg, P.. (2022). A deep convolutional neural network that is invariant to time rescaling. Proceedings of the 39th International Conference on Machine Learning, in Proceedings of Machine Learning Research 162:9729-9738 Available from https://proceedings.mlr.press/v162/jacques22a.html.

Related Material