Learning nonlinear dynamic models

By J. Langford, Ruslan Salakhutdinov, Tong Zhang. 2009

50 citations recorded by the source corpus.

We present a novel approach for learning nonlinear dynamic models, which leads to a new set of tools capable of solving problems that are otherwise difficult. We provide theory showing this new approach is consistent for models with long range structure, and apply the approach to motion capture and high-dimensional video data, yielding results superior to standard alternatives.

Subfields: Artificial Intelligence, Signal Processing, Time-Series & Forecasting

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

In Signal Processing

In Time-Series & Forecasting

Citations in the corpus

Embed to Control: A Locally Linear Latent Dynamics Model for Control from Raw Images (2015, background): “[36] for a recent discussion and a general formalization of this problem.”
Deep Kalman Filters (2015, methodology): “Langford et al. (2009) adopt a different approach to learn nonlinear dynamical systems using black-box classiﬁers.”
Improving Multi-Step Prediction of Learned Time Series Models (2015, methodology): “If the predictive model is differentiable, one can apply “backpropagation-through-time” (Werbos 1990; Langford, Salakhutdinov, and Zhang 2009).”
Learning Deep Hidden Nonlinear Dynamics from Aggregate Data (2018, background): “However, these models and their variants (Langford et al., 2009; Hefny et al., 2015) require individual-level trajectories, which may not be available, as was mentioned earlier.”
Learning to Filter with Predictive State Inference Machines (2015, methodology): “…reduces the problem of learning graphical models to solving a set of classification or regression problems, where the learned classifiers mimic message passing procedures that output marginal distributions for the nodes in the model (Langford et al., 2009; Ross et al., 2011b; Bagnell et al., 2010).”
A Separation Principle for Control in the Age of Deep Learning (2017, background): “More directly related to our approach, [26] suggest that, rather than ﬁnding a (generally complex) hidden state, we could focus on ﬁnding a statistic x t of the past data y t that separates past data from future predictions.”
Predictive-State Decoders: Encoding the Future into Recurrent Networks (2017, methodology): “[45] extended this setup to create a practical algorithm, Predictive-State Inference Machines (PSIMs) [44, 45, 54], based on the concept of inference machines [31, 40].”