Learning the Structure of Deep Sparse Graphical Models

Deep belief networks are a powerful way to model complex probability distributions. However, learning the structure of a belief network, particularly one with hidden units, is difficult. The Indian buffet process has been used as a nonparametric Bayesian prior on the directed structure of a belief network with a single infinitely wide hidden layer. In this paper, we introduce the cascading Indian buffet process (CIBP), which provides a nonparametric prior on the structure of a layered, directed belief network that is unbounded in both depth and width, yet allows tractable inference. We use the CIBP prior with the nonlinear Gaussian belief network so each unit can additionally vary its behavior between discrete and continuous representations. We provide Markov chain Monte Carlo algorithms for inference in these belief networks and explore the structures learned on several image data sets.

The internal structure of igniting turbulent sprays as revealed by complex chemistry DNS

A parametric study of spark ignition in a uniform monodisperse turbulent spray is performed with complex chemistry three-dimensional Direct Numerical Simulations in order to improve the understanding of the structure of the ignition kernel. The heat produced by the kernel increases with the amount of fuel evaporated inside the spark volume. Moreover, the heat sink by evaporation is initially higher than the heat release and can have a negative effect on ignition. With the sprays investigated, heat release occurs over a large range of mixture fractions, being high within the nominal flammability limits and finite but low below the lean flammability limit. The burning of very lean regions is attributed to the diffusion of heat and species from regions of high heat release, and from the spark, to lean regions. Two modes of spray ignition are reported. With a relatively dilute spray, nominally flammable material exists only near the droplets. Reaction zones are created locally near the droplets and have a non-premixed character. They spread from droplet to droplet through a very lean interdroplet spacing. With a dense spray, the hot spark region is rich due to substantial evaporation but the cold region remains lean. In between, a large surface of flammable material is generated by evaporation. Ignition occurs there and a large reaction zone propagates from the rich burned region to the cold lean region. This flame is wrinkled due to the stratified mixture fraction field and evaporative cooling. In the dilute spray, the reaction front curvature pdf contains high values associated with single droplet combustion, while in the dense spray, the curvature is lower and closer to the curvature associated with gaseous fuel ignition kernels. © 2011 The Combustion Institute.

Recovering 3D Structure of Poorly Textured Infrastructure Scenes Using Point and Line Features

Most of the existing automated machine vision-based techniques for as-built documentation of civil infrastructure utilize only point features to recover the 3D structure of a scene. However it is often the case in man-made structures that not enough point features can be reliably detected (e.g. buildings and roofs); this can potentially lead to the failure of these techniques. To address the problem, this paper utilizes the prominence of straight lines in infrastructure scenes. It presents a hybrid approach that benefits from both point and line features. A calibrated stereo set of video cameras is used to collect data. Point and line features are then detected and matched across video frames. Finally, the 3D structure of the scene is recovered by finding 3D coordinates of the matched features. The proposed approach has been tested on realistic outdoor environments and preliminary results indicate its capability to deal with a variety of scenes.