Sajid M. Siddiqi

Closing the learning-planning loop with predictive state representations

A central problem in artificial intelligence is to choose actions to maximize reward in a partially observable, uncertain environment. To do so, we must learn an accurate environment model, and then plan to maximize reward. Unfortunately, learning algorithms often recover a model that is too inaccurate to support planning or too large and complex for planning to succeed; or they require excessive prior domain knowledge or fail to provide guarantees such as statistical consistency. To address this gap, we propose a novel algorithm which provably learns a compact, accurate model directly from sequences of action-observation pairs. We then evaluate the learner by closing the loop from observations to actions. In more detail, we present a spectral algorithm for learning a predictive state representation (PSR), and evaluate it in a simulated, vision-based mobile robot planning task, showing that the learned PSR captures the essential features of the environment and enables successful and efficient planning. Our algorithm has several benefits which have not appeared together in any previous PSR learner: it is computationally efficient and statistically consistent; it handles high-dimensional observations and long time horizons; and, our close-the-loop experiments provide an end-to-end practical test.

Fast inference and learning in large-state-space HMMs

For Hidden Markov Models (HMMs) with fully connected transition models, the three fundamental problems of evaluating the likelihood of an observation sequence, estimating an optimal state sequence for the observations, and learning the model parameters, all have quadratic time complexity in the number of states. We introduce a novel class of non-sparse Markov transition matrices called Dense-Mostly-Constant (DMC) transition matrices that allow us to derive new algorithms for solving the basic HMM problems in sub-quadratic time. We describe the DMC HMM model and algorithms and attempt to convey some intuition for their usage. Empirical results for these algorithms show dramatic speedups for all three problems. In terms of accuracy, the DMC model yields strong results and outperforms the baseline algorithms even in domains known to violate the DMC assumption.

Automatic state discovery for unstructured audio scene classification

In this paper we present a novel scheme for unstructured audio scene classification that possesses three highly desirable and powerful features: autonomy, scalability, and robustness. Our scheme is based on our recently introduced machine learning algorithm called Simultaneous Temporal And Contextual Splitting (STACS) that discovers the appropriate number of states and efficiently learns accurate Hidden Markov Model (HMM) parameters for the given data. STACS-based algorithms train HMMs up to five times faster than Baum-Welch, avoid the overfitting problem commonly encountered in learning large state-space HMMs using Expectation Maximization (EM) methods such as Baum-Welch, and achieve superior classification results on a very diverse dataset with minimal pre-processing. Furthermore, our scheme has proven to be highly effective for building real-world applications and has been integrated into a commercial surveillance system as an event detection component.

Approximate Kalman filters for embedding author-word co-occurrence data over time

We address the problem of embedding entities into Euclidean space over time based on co-occurrence data. We extend the CODE model of [1] to a dynamic setting. This leads to a non-standard factored state space model with real-valued hidden parent nodes and discrete observation nodes. We investigate the use of variational approximations applied to the observation model that allow us to formulate the entire dynamic model as a Kalman filter. Applying this model to temporal co-occurrence data yields posterior distributions of entity coordinates in Euclidean space that are updated over time. Initial results on per-year co-occurrences of authors and words in the NIPS corpus and on synthetic data, including videos of dynamic embeddings, seem to indicate that the model results in embeddings of co-occurrence data that are meaningful both temporally and contextually.