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.

INFLUENCE OF ATMOSPHERIC HUMIDITY ON THE ABRASIVE WEAR OF METALS.

A pin-on-disc apparatus has been used to obtain continuous simultaneous measurements of the wear and friction (sliding force) behaviour of metals on bonded silicon carbide abrasive paper under conditions of controlled humidity. Iron, mild steel, and copper exhibit qualitatively similar wear behaviour; the wear rate decreases progressively with the number of passes over the same track. In contrast, the wear rate of titanium remains constant. Variation in atmospheric humidity has little effect on the wear rates of copper or titanium, although a slight effect was found in mild steel and iron. Refs.

INFLUENCE OF ATMOSPHERIC COMPOSITION ON THE ABRASIVE WEAR OF TITANIUM AND Ti-6Al-4V.

A pin-on-disc apparatus has been used to investigate the wear and friction (sliding force) behavior of metals on bonded silicon carbide and alumina papers under conditions of controlled atmospheric composition. The wear rates of both commercial purity titanium and the alloy Ti-6%Al-4%V tested in air were found to remain constant with time, in contrast with the behavior of other metals tested under similar conditions, which exhibited a progressive decrease in wear rate with increasing number of passes along the same track. It is proposed that the concentration of interstitial nitrogen and oxygen in the worn metal surface, which largely determines its mechanical properties, strongly influences both the ductility of the abraded material and the force of adhesion between the metal and the abrasive particles. Parallels are drawn between abrasive wear and machining to illustrate the importance of oxygen at the interface between workpiece and tool surfaces.