Topology, Convergence, and Reconstruction of Predictive States

Predictive equivalence in discrete stochastic processes have been applied with great success to identify randomness and structure in statistical physics and chaotic dynamical systems and to inferring hidden Markov models. We examine the conditions under which they can be reliably reconstructed from time-series data, showing that convergence of predictive states can be achieved from empirical samples in the weak topology of measures. Moreover, predictive states may be represented in Hilbert spaces that replicate the weak topology. We mathematically explain how these representations are particularly beneficial when reconstructing high-memory processes and connect them to reproducing kernel Hilbert spaces.

[1]  Petr Kůrka,et al.  Topological and symbolic dynamics , 2003 .

[2]  Le Song,et al.  Hilbert Space Embeddings of Hidden Markov Models , 2010, ICML.

[3]  James Odell,et al.  Between order and chaos , 2011, Nature Physics.

[4]  Herbert Jaeger,et al.  Links between multiplicity automata, observable operator models and predictive state representations: a unified learning framework , 2015, J. Mach. Learn. Res..

[5]  M. Gu,et al.  Practical Unitary Simulator for Non-Markovian Complex Processes. , 2017, Physical review letters.

[6]  B. Jessen.  The theory of integration in a space of an infinite number of dimensions , 1934 .

[7]  James P. Crutchfield,et al.  Structure and Randomness of Continuous-Time, Discrete-Event Processes , 2017, ArXiv.

[8]  James P. Crutchfield,et al.  Optimized Bacteria are Environmental Prediction Engines , 2018, Physical review. E.

[9]  James P. Crutchfield,et al.  Predictive Rate-Distortion for Infinite-Order Markov Processes , 2016 .

[10]  James P. Crutchfield,et al.  Bayesian Structural Inference for Hidden Processes , 2013, Physical review. E, Statistical, nonlinear, and soft matter physics.

[11]  Le Song,et al.  Kernel Bayes' rule: Bayesian inference with positive definite kernels , 2013, J. Mach. Learn. Res..

[12]  Adam Rupe,et al.  Local Causal States and Discrete Coherent Structures , 2018, Chaos.

[13]  Susanne Still,et al.  Optimal causal inference: estimating stored information and approximating causal architecture. , 2007, Chaos.

[14]  James P. Crutchfield,et al.  Asymptotic Synchronization for Finite-State Sources , 2010, ArXiv.

[15]  Liang Zhao,et al.  On the Inclusion Relation of Reproducing Kernel Hilbert Spaces , 2011, ArXiv.

[16]  M. Rao.  Conditional measures and applications , 1993 .

[17]  J. Crutchfield,et al.  Thermal Efficiency of Quantum Memory Compression , 2019, Physical review letters.

[18]  Byron Boots,et al.  Hilbert Space Embeddings of Predictive State Representations , 2013, UAI.

[19]  A. B. Boyd,et al.  Maxwell Demon Dynamics: Deterministic Chaos, the Szilard Map, and the Intelligence of Thermodynamic Systems. , 2015, Physical review letters.

[20]  Kristina Lisa Shalizi,et al.  Pattern Discovery in Time Series, Part I: Theory, Algorithm, Analysis, and Convergence , 2002 .

[21]  Michael R. James,et al.  Learning and discovery of predictive state representations in dynamical systems with reset , 2004, ICML.

[22]  Karoline Wiesner,et al.  Quantum mechanics can reduce the complexity of classical models , 2011, Nature Communications.

[23]  Herbert Jaeger,et al.  Observable Operator Models for Discrete Stochastic Time Series , 2000, Neural Computation.

[24]  O. Kallenberg.  Foundations of Modern Probability , 2021, Probability Theory and Stochastic Modelling.

[25]  Shun-ichi Amari,et al.  Identifiability of hidden Markov information sources and their minimum degrees of freedom , 1992, IEEE Trans. Inf. Theory.

[26]  Bernhard Schölkopf,et al.  Kernel Mean Embedding of Distributions: A Review and Beyonds , 2016, Found. Trends Mach. Learn..

[27]  Adam Rupe,et al.  DisCo: Physics-Based Unsupervised Discovery of Coherent Structures in Spatiotemporal Systems , 2019, 2019 IEEE/ACM Workshop on Machine Learning in High Performance Computing Environments (MLHPC).

[28]  Ingo Steinwart.  Reproducing Kernel Hilbert Spaces Cannot Contain all Continuous Functions on a Compact Metric Space , 2020, ArXiv.

[29]  James P. Crutchfield,et al.  Occam’s Quantum Strop: Synchronizing and Compressing Classical Cryptic Processes via a Quantum Channel , 2015, Scientific Reports.

[30]  James P. Crutchfield,et al.  Chaotic Crystallography: How the physics of information reveals structural order in materials , 2014, ArXiv.

[31]  Jeffrey D. Ullman,et al.  Introduction to Automata Theory, Languages and Computation , 1979 .

[32]  James P. Crutchfield,et al.  Divergent Predictive States: The Statistical Complexity Dimension of Stationary, Ergodic Hidden Markov Processes , 2021, Chaos.

[33]  N. Aronszajn.  Theory of Reproducing Kernels. , 1950 .

[34]  J. Crutchfield,et al.  Statistical complexity of simple one-dimensional spin systems , 1997, cond-mat/9702191.

[35]  J. Crutchfield,et al.  Discovering Causal Structure with Reproducing-Kernel Hilbert Space ε-Machines , 2020, Chaos.

[36]  Daniel Ray Upper,et al.  Theory and algorithms for hidden Markov models and generalized hidden Markov models , 1998 .

[37]  Bernhard Schölkopf,et al.  Hilbert Space Embeddings and Metrics on Probability Measures , 2009, J. Mach. Learn. Res..

[38]  James P. Crutchfield,et al.  Measurement-induced randomness and structure in controlled qubit processes. , 2019, Physical review. E.

[39]  James P. Crutchfield,et al.  Functional Thermodynamics of Maxwellian Ratchets: Constructing and Deconstructing Patterns, Randomizing and Derandomizing Behaviors , 2020 .

[40]  Alexander J. Smola,et al.  Hilbert space embeddings of conditional distributions with applications to dynamical systems , 2009, ICML '09.

[41]  Adam Rupe,et al.  Spacetime Autoencoders Using Local Causal States , 2020, ArXiv.