This paper presents a novel algorithm which generates three-dimensional face point trajectories for a given speech le with or without its text. The proposed algorithm rst employs an oo-line training phase. In this phase, recorded face point trajectories along with their speech data and phonetic labels are used to generate phonetic codebooks. These codebooks consist of both acoustic and visual f...

Evolutionary algorithms EAs are stochas tic population based algorithms inspired by the natural processes of recombination mu tation and selection EAs are often em ployed as optimum seeking techniques A for mal framework for EAs is proposed in which evolutionary operators are viewed as map pings from parameter spaces to spaces of ran dom functions Formal de nitions within this framework capture...

We present a general framework for defining priors on model structure and sampling from the posterior using the Metropolis-Hastings algo­ rithm. The key ideas are that structure priors are defined via a probability tree and that the pro­ posal distribution for the Metropolis-Hastings al­ gorithm is defined using the prior, thereby defin­ ing a cheaply computable acceptance probabil­ ity. We hav...

We present probabilistic neural programs, a framework for program induction that 1 permits flexible specification of both a computational model and inference algo2 rithm while simultaneously enabling the use of deep neural networks. Probabilistic 3 neural programs combine a computation graph for specifying a neural network with 4 an operator for weighted nondeterministic choice. Thus, a program...

rithm that is provably sound and complete across a large class of planning domains. It plans for tasks in the same order that they will later be executed, and thus, it knows the current world state at each step of the planning process. SHOP takes advantage of this knowledge by allowing a high degree of expressive power in its knowledge bases. For example, SHOP’s preconditions can include logica...

We present a new algorithm for inferring the home locations of Twitter users at different granularities, such as city, state, or time zone, using the content of their tweets and their tweeting behavior. Unlike existing approaches, our algo rithm uses an ensemble of statistical and heuristic classifiers to predict locations. We find that a hierarchical classifica tion approach can improve predic...

In this paper a particle gradient evolutionary algorithm is presented for solving complex single-objective optimization problems based on statistical mechanics theory, the principle of gradient descending, and the law of evolving chance ascending of particles. Numerical experiments show that we can easily solve complex single-objective optimiz ation problems that are difficult to solve by using...

In this paper we study geometrical structures on the manifold of FIR lters and their application to multichannel blind deconvolution First we introduce the Lie group and Riemannian metric to the manifold of FIR lters Then we derive the natu ral gradient on the manifold using the isometry of the Riemannian metric Using the natural gradient we present a novel learning algorithm for blind deconvol...

We propose a rigorous study of an ierative maximization algo rithm introduced by Geyer and Thompson for maximum likelihood estimation of Markov random elds One step of the algorithm consists in a Monte Carlo approximation of the likelihood followed by a local maximization in the neigh borhood of the current parameter We study convergence properties of the induced process and bound the computati...

An algorithm is resented for con8tructing a quadtree from the array repre+@&oa of a binary u. %h e algorithm examines each pixel in the w once aasd only once. In addition, aa the tree ie con&u&d, only m&iial &ed nodea are ever cd. Thua the algorithm never requirea temporary nodes. The execution time of the B@-rithm is equal to the number of pixel43 in the we. The amount of sp&oe, in 84$& tion t...