Exploring predictive states via Cantor embeddings and Wasserstein distance
Open Access
- 1 December 2022
- journal article
- research article
- Published by AIP Publishing in Chaos: An Interdisciplinary Journal of Nonlinear Science
- Vol. 32 (12) , 123115
Abstract
Predictive states for stochastic processes are a nonparametric and interpretable construct with relevance across a multitude of modeling paradigms. Recent progress on the self-supervised reconstruction of predictive states from time-series data focused on the use of reproducing kernel Hilbert spaces. Here, we examine how Wasserstein distances may be used to detect predictive equivalences in symbolic data. We compute Wasserstein distances between distributions over sequences (“predictions”) using a finite-dimensional embedding of sequences based on the Cantor set for the underlying geometry. We show that exploratory data analysis using the resulting geometry via hierarchical clustering and dimension reduction provides insight into the temporal structure of processes ranging from the relatively simple (e.g., generated by finite-state hidden Markov models) to the very complex (e.g., generated by infinite-state indexed grammars).Keywords
All Related Versions
Funding Information
- Army Research Office (W911NF-18-1-0028)
- Army Research Office (W911NF-21-1-0048)
- Templeton World Charity Foundation (TWCF0570)
- Foundational Questions Institute (FQXI-RFP-CPW-2007)
- Templeton World Charity Foundation (TWCF0440)
This publication has 15 references indexed in Scilit:
- Statistical signatures of structural organization: The case of long memory in renewal processesPhysics Letters A, 2016
- Time resolution dependence of information measures for spiking neurons: scaling and universalityFrontiers in Computational Neuroscience, 2015
- Chaotic crystallography: how the physics of information reveals structural order in materialsCurrent Opinion in Chemical Engineering, 2015
- Bayesian structural inference for hidden processesPhysical Review E, 2014
- Optimal causal inference: Estimating stored information and approximating causal architectureChaos: An Interdisciplinary Journal of Nonlinear Science, 2010
- Thermodynamic depth of causal states: Objective complexity via minimal representationsPhysical Review E, 1999
- The calculi of emergence: computation, dynamics and inductionPhysica D: Nonlinear Phenomena, 1994
- Stochastic context-free grammers for tRNA modelingNucleic Acids Research, 1994
- Inferring statistical complexityPhysical Review Letters, 1989
- The Future of Data AnalysisThe Annals of Mathematical Statistics, 1962