Independent Component Discovery in Temporal Count Data

Alexandre Chaussard, Anna Bonnet, Sylvain Le Corff
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:13213-13246, 2026.

Abstract

Advances in data collection are producing growing volumes of temporal count observations, making adapted modeling increasingly necessary. In this work, we introduce a generative framework for independent component analysis of temporal count data, combining regime-adaptive dynamics with Poisson log-normal emissions. The model identifies disentangled components with regime-dependent contributions, enabling representation learning and perturbations analysis. Notably, we establish the identifiability of the model, supporting principled interpretation. To learn the parameters, we propose an efficient amortized variational inference procedure. Experiments on simulated data evaluate recovery of the mixing function and latent sources across diverse settings, while real-world applications to gut microbiome and climate datasets reveal co-variation patterns and regime shifts consistent with domain-specific knowledge.

Cite this Paper

BibTeX
@InProceedings{pmlr-v306-chaussard26a, title = {Independent Component Discovery in Temporal Count Data}, author = {Chaussard, Alexandre and Bonnet, Anna and Le Corff, Sylvain}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {13213--13246}, year = {2026}, editor = {Zhang, Tong and Dudik, Miroslav and Jaggi, Martin and Agarwal, Alekh and Li, Sharon and Schuurmans, Dale and Zhu, Jerry and Berkenkamp, Felix and Dong, Hanze and Bietti, Alberto}, volume = {306}, series = {Proceedings of Machine Learning Research}, month = {06--11 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v306/main/assets/chaussard26a/chaussard26a.pdf}, url = {https://proceedings.mlr.press/v306/chaussard26a.html}, abstract = {Advances in data collection are producing growing volumes of temporal count observations, making adapted modeling increasingly necessary. In this work, we introduce a generative framework for independent component analysis of temporal count data, combining regime-adaptive dynamics with Poisson log-normal emissions. The model identifies disentangled components with regime-dependent contributions, enabling representation learning and perturbations analysis. Notably, we establish the identifiability of the model, supporting principled interpretation. To learn the parameters, we propose an efficient amortized variational inference procedure. Experiments on simulated data evaluate recovery of the mixing function and latent sources across diverse settings, while real-world applications to gut microbiome and climate datasets reveal co-variation patterns and regime shifts consistent with domain-specific knowledge.} }
Endnote
%0 Conference Paper %T Independent Component Discovery in Temporal Count Data %A Alexandre Chaussard %A Anna Bonnet %A Sylvain Le Corff %B Proceedings of the 43rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2026 %E Tong Zhang %E Miroslav Dudik %E Martin Jaggi %E Alekh Agarwal %E Sharon Li %E Dale Schuurmans %E Jerry Zhu %E Felix Berkenkamp %E Hanze Dong %E Alberto Bietti %F pmlr-v306-chaussard26a %I PMLR %P 13213--13246 %U https://proceedings.mlr.press/v306/chaussard26a.html %V 306 %X Advances in data collection are producing growing volumes of temporal count observations, making adapted modeling increasingly necessary. In this work, we introduce a generative framework for independent component analysis of temporal count data, combining regime-adaptive dynamics with Poisson log-normal emissions. The model identifies disentangled components with regime-dependent contributions, enabling representation learning and perturbations analysis. Notably, we establish the identifiability of the model, supporting principled interpretation. To learn the parameters, we propose an efficient amortized variational inference procedure. Experiments on simulated data evaluate recovery of the mixing function and latent sources across diverse settings, while real-world applications to gut microbiome and climate datasets reveal co-variation patterns and regime shifts consistent with domain-specific knowledge.
APA
Chaussard, A., Bonnet, A. & Le Corff, S.. (2026). Independent Component Discovery in Temporal Count Data. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:13213-13246 Available from https://proceedings.mlr.press/v306/chaussard26a.html.

Related Material