Observable Operator Models for Discrete Stochastic Time Series

By Herbert Jaeger. 2000

246 citations recorded by the source corpus.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

Citations in the corpus

Kernel Mean Embedding of Distributions: A Review and Beyonds (2016, background): “For intractable models, Jebara et al. (2004) suggested approximating (3.”
Closing the learning-planning loop with predictive state representations (2009, background): “And, many successful approximate planning techniques for POMDPs can be used to plan in PSRs and OOMs with minimal adjustment.”
Tensor decompositions for learning latent variable models (2012, methodology): “Notably, another class of methods, based on subspace identification (Overschee and Moor, 1996) and observable operator models/multiplicity automata (Schützenberger, 1961; Jaeger, 2000; Littman et al., 2001), have been proposed for a number of latent variable models.”
Reinforcement Learning: An Introduction (1998, background): “OOMs were introduced by Jaeger (1997, 1998, 2000).”
The Grand Challenge of Predictive Empirical Abstract Knowledge (2009, background): “The world show in t e figure is sufficiently small that complete knowledge of it was represented and learned using a temporal-difference (TD) net ork (Sutton & Tanner, 2005), a form of predictive state representation (Littman, Sutton & Singh, 2002; Rosencrantz, Gordon & Thrun, 2004; Jaeger, 2000).”
Computational Mechanics: Pattern and Prediction, Structure and Simplicity (1999, unknown): “D, linking an algebraic operation with concate-nations of strings, is analogous in a rough way, as are the \observable operator models" of Ref.”
Learning Predictive State Representations (2003, background): “Analogous demonstrations have been made for the diversity representation (Rivest & Schapire, 1994) and the observable operator representation (Jaeger, 2000).”
A spectral algorithm for learning Hidden Markov Models (2008, methodology): “The following lemma is straightforward to verify (see [Jaeger,2000, Even-Dar et al., 2007]).”
A Spectral Algorithm for Learning Hidden Markov Models (2009, methodology): “In fact, the former work by [Jae00] provides a noniterative algorithm for learning HMMs, with an asymptotic analysis.”
Sample-Efficient Reinforcement Learning for POMDPs with Linear Function Approximations (2022, methodology): “Their proposed algorithm can attain the optimal policy without estimating the exact model, but an observable component (Jaeger, 2000; Hsu et al., 2012), which is the same for our algorithm design, while only applies to tabular POMDPs.”
A Method of Moments for Mixture Models and Hidden Markov Models (2012, background): “Similar observable operators were previously used to characterize multiplicity automata (Sch¨utzenberger, 1961; Jaeger, 2000) and, more recently, for learning discrete HMMs (via an operator parameterization) (Hsu et al., 2009).”
Approximate PCFG Parsing Using Tensor Decomposition (2013, background): “A similar observation has been made for latent-variable PCFGs (Cohen et al., 2012) and hidden Markov models, where only matrix multiplication is required (Jaeger, 2000).”
An Online Spectral Learning Algorithm for Partially Observable Nonlinear Dynamical Systems (2011, background): “(Elements of H are called indicative events (Jaeger 2000).)”
Learning predictive models of a depth camera & manipulator from raw execution traces (2014, background): “In the machine learning community, where probabilistic models play a central role, the identification problem is typically concerned with learning the parameters of a controlled stochastic process that generated the observations [11], [25], [26].”
Generating Text with Recurrent Neural Networks (2011, methodology): “…large, then the storage required for the tensor W The above model, applied to discrete inputs represented with their 1-of-M encodings, is the nonlinear version of the Observable Operator Model (OOM; Jaeger, 2000) whose linear nature makes it closely related to an HMM in terms of expressive power.”
Blind Construction of Optimal Nonlinear Recursive Predictors for Discrete Sequences (2004, methodology): “Suffix-tree methods can induce POMDPs from data [23, 27, 28], and we believe CSSR can be adapted to reconstruct POMDPs.”