Current statistical models for structured prediction make simplifying assumptions about the underlying output graph structure, such as assuming a low-order Markov chain, because exact inference becomes intractable as the tree-width of the underlying graph increases. Approximate inference algorithms, on the other hand, force one to trade off representational power with computational efficiency. In this paper, we propose two new types of probabilistic graphical models, large margin Boltzmann machines (LMBMs) and large margin sigmoid belief networks (LMSBNs), for structured prediction. LMSBNs in particular allow a very fast inference algorithm for arbitrary graph structures that runs in polynomial time with a high probability. This probability is data-distribution dependent and is maximized in learning. The new approach overcomes the representation-efficiency trade-off in previous models and allows fast structured prediction with complicated graph structures. We present results from applying a fully connected model to multi-label scene classification and demonstrate that the proposed approach can yield significant performance gains over current state-of-the-art methods.

On interactive surfaces, an accurate system calibration is crucial for a precise user interaction. Today, geometric distortions are eliminated by a static calibration. However, this calibration is specific to a user’s posture, and parallax distortions occur if this changes (i.e. if the user moves or if multiple users take turns). Within this paper, we describe an approach to model automatic online re-calibration to cope with changing viewpoints by using Partially Observable Markov Decision Processes (POMDP). Hereby, the viewpoint is stochastically deducted from the precision of user interactions on the surface. To enable the implementation on embedded systems, a small model is defined using states and observations, which are formulated relative to the current assumed viewpoint. We show the structure of a family of models, that can be generated automatically based on the user’s position probability and pointing accuracy.

The correct recognition of behaviours based on sensor observations in a smart home is a challenging problem; the sensor observations themselves can be noisy, and the pattern activity seen for a behaviour is rarely identical for different occurrences of the behaviour. For this reason, probabilistic methods such as Hidden Markov Models are preferred over symbolic reasoning approaches. However, these models do not deal well with interleaved behaviours, nor do they allow small variations in behaviour to be detected as abnormal, although this might be useful for the smart home, since changes in ingrained habit could be early signs of illness. We propose methods for using Allen's temporal relations in order to solve these problems, and demonstrate how they can be used to recognise the interleaving of different behaviours, as well as to reason about behaviours that are frequently seen together, and therefore form a behavioural pattern or habit. In this way we have been able to extend our behaviour recognition system to recognise unusual presentations of behaviours.

We present an exact dynamic programming solution for a finite-horizon decentralized two-player Markov decision process, where player 1 only has access to its own states, while player 2 has access to both player’s states but cannot affect player 1’s states. The solution is obtained by solving several centralized partially-observable Markov decision processes. We then conclude with several computational examples.

Many probabilistic models introduce strong dependencies between variables using a latent multivariate Gaussian distribution or a Gaussian process. We present a new Markov chain Monte Carlo algorithm for performing inference in models with multivariate Gaussian priors. Its key properties are: 1) it has simple, generic code applicable to many models, 2) it has no free parameters, 3) it works well for a variety of Gaussian process based models. These properties make our method ideal for use while model building, removing the need to spend time deriving and tuning updates for more complex algorithms.

Adaptive control problems are notoriously difficult to solve even in the presence of plant-specific controllers. One way to by-pass the intractable computation of the optimal policy is to restate the adaptive control as the minimization of the relative entropy of a controller that ignores the true plant dynamics from an informed controller. The solution is given by the Bayesian control rule-a set of equations characterizing a stochastic adaptive controller for the class of possible plant dynamics. Here, the Bayesian control rule is applied to derive BCR-MDP, a controller to solve undiscounted Markov decision processes with finite state and action spaces and unknown dynamics. In particular, we derive a non-parametric conjugate prior distribution over the policy space that encapsulates the agent's whole relevant history and we present a Gibbs sampler to draw random policies from this distribution. Preliminary results show that BCR-MDP successfully avoids sub-optimal limit cycles due to its built-in mechanism to balance exploration versus exploitation.

Programming by demonstration systems have long attempted to make it possible for people to program computers without writing code. However, while these systems have resulted in many publications in AI venues, none of the technologies have yet achieved widespread.adoption. Usability remains a critical barrier to their success. On the basis of lessons learned from three different programming by demonstration systems, we present a set of guidelines to consider when designing usable AI-based systems.

A score function induced by a generative model of the data can provide a feature vectorof a fixed dimension for each data sample. Data samples themselves may be of differing lengths (e.g., speech segments, or other sequence data), but as a score function is based on the properties of the data generation process, it produces a fixed-length vector in a highly informative space, typically referred to as a "score space". Discriminative classifiers have been shown to achieve higher performance in appropriately chosen score spaces than is achievable by either the corresponding generative likelihood-based classifiers, or the discriminative classifiers usingstandard feature extractors. In this paper, we present a novel score space that exploits the free energy associated with a generative model. The resulting freeenergy score space (FESS) takes into account latent structure of the data at various levels, and can be trivially shown to lead to classification performance that at least matches the performance of the free energy classifier based on the same generative model, and the same factorization of the posterior. We also show that in several typical vision and computational biology applications the classifiers optimized in FESS outperform the corresponding pure generative approaches, as well as a number of previous approaches to combining discriminating and generative models.

We present a nonparametric Bayesian method for texture learning and synthesis. A texture image is represented by a 2D-Hidden Markov Model (2D-HMM) where the hidden states correspond to the cluster labeling of textons and the transition matrix encodes their spatial layout (the compatibility between adjacent textons). 2D-HMM is coupled with the Hierarchical Dirichlet process (HDP) which allows the number of textons and the complexity of transition matrix grow as the input texture becomes irregular. The HDP makes use of Dirichlet process prior which favors regular textures by penalizing the model complexity. This framework (HDP-2D-HMM) learns the texton vocabulary and their spatial layout jointly and automatically. The HDP-2D-HMM results in a compact representation of textures which allows fast texture synthesis with comparable rendering quality over the state-of-the-art image-based rendering methods. We also show that HDP-2D-HMM can be applied to perform image segmentation and synthesis.

Neural activity is non-stationary and varies across time. Hidden Markov Models (HMMs) have been used to track the state transition among quasi-stationary discrete neural states. Within this context, independent Poisson models have been used for the output distribution of HMMs; hence, the model is incapable of tracking the change in correlation without modulating the firing rate. To achieve this, we applied a multivariate Poisson distribution with correlation terms for the output distribution of HMMs. We formulated a Variational Bayes (VB) inference for the model. The VB could automatically determine the appropriate number of hidden states and correlation types while avoiding the overlearning problem. We developed an efficient algorithm for computing posteriors using the recursive relationship of a multivariate Poisson distribution. We demonstrated the performance of our method on synthetic data and a real spike train recorded from a songbird.