TY - GEN
T1 - Rate-Invariant Autoencoding of Time-Series
AU - Koneripalli, Kaushik
AU - Lohit, Suhas
AU - Anirudh, Rushil
AU - Turaga, Pavan
N1 - Funding Information:
This work was supported by NSF grant 1617999. The authors would like to thank Ankita Shukla for useful discussions, and NVIDIA Corporation for donating a Titan Xp GPU which was used for some experiments in this paper.
PY - 2020/5
Y1 - 2020/5
N2 - For time-series classification and retrieval applications, an important requirement is to develop representations/metrics that are robust to re-parametrization of the time-axis. Temporal re-parametrization as a model can account for variability in the underlying generative process, sampling rate variations, or plain temporal mis-alignment. In this paper, we extend prior work in disentangling latent spaces of autoencoding models, to design a novel architecture to learn rate-invariant latent codes in a completely unsupervised fashion. Unlike conventional neural network architectures, this method allows to explicitly disentangle temporal parameters in the form of order-preserving diffeomorphisms with respect to a learnable template. This makes the latent space more easily interpretable. We show the efficacy of our approach on a synthetic dataset and a real dataset for hand action-recognition.
AB - For time-series classification and retrieval applications, an important requirement is to develop representations/metrics that are robust to re-parametrization of the time-axis. Temporal re-parametrization as a model can account for variability in the underlying generative process, sampling rate variations, or plain temporal mis-alignment. In this paper, we extend prior work in disentangling latent spaces of autoencoding models, to design a novel architecture to learn rate-invariant latent codes in a completely unsupervised fashion. Unlike conventional neural network architectures, this method allows to explicitly disentangle temporal parameters in the form of order-preserving diffeomorphisms with respect to a learnable template. This makes the latent space more easily interpretable. We show the efficacy of our approach on a synthetic dataset and a real dataset for hand action-recognition.
KW - Rate-invariance
KW - autoencoder
KW - deep learning
KW - neural networks
KW - time-series
UR - http://www.scopus.com/inward/record.url?scp=85089245703&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85089245703&partnerID=8YFLogxK
U2 - 10.1109/ICASSP40776.2020.9053983
DO - 10.1109/ICASSP40776.2020.9053983
M3 - Conference contribution
AN - SCOPUS:85089245703
T3 - ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings
SP - 3732
EP - 3736
BT - 2020 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2020 - Proceedings
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 2020 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2020
Y2 - 4 May 2020 through 8 May 2020
ER -