Reformulation of Ensemble Averages via Coordinate Mapping.

A general framework is established for reformulation of the ensemble averages commonly encountered in statistical mechanics. This "mapped-averaging" scheme allows approximate theoretical results that have been derived from statistical mechanics to be reintroduced into the underlying formalism, yielding new ensemble averages that represent exactly the error in the theory. The result represents a distinct alternative to perturbation theory for methodically employing tractable systems as a starting point for describing complex systems. Molecular simulation is shown to provide one appealing route to exploit this advance. Calculation of the reformulated averages by molecular simulation can proceed without contamination by noise produced by behavior that has already been captured by the approximate theory. Consequently, accurate and precise values of properties can be obtained while using less computational effort, in favorable cases, many orders of magnitude less. The treatment is demonstrated using three examples: (1) calculation of the heat capacity of an embedded-atom model of iron, (2) calculation of the dielectric constant of the Stockmayer model of dipolar molecules, and (3) calculation of the pressure of a Lennard-Jones fluid. It is observed that improvement in computational efficiency is related to the appropriateness of the underlying theory for the condition being simulated; the accuracy of the result is however not impacted by this. The framework opens many avenues for further development, both as a means to improve simulation methodology and as a new basis to develop theories for thermophysical properties.

A computational visual saliency model based on statistics and machine learning.

Identifying the type of stimuli that attracts human visual attention has been an appealing topic for scientists for many years. In particular, marking the salient regions in images is useful for both psychologists and many computer vision applications. In this paper, we propose a computational approach for producing saliency maps using statistics and machine learning methods. Based on four assumptions, three properties (Feature-Prior, Position-Prior, and Feature-Distribution) can be derived and combined by a simple intersection operation to obtain a saliency map. These properties are implemented by a similarity computation, support vector regression (SVR) technique, statistical analysis of training samples, and information theory using low-level features. This technique is able to learn the preferences of human visual behavior while simultaneously considering feature uniqueness. Experimental results show that our approach performs better in predicting human visual attention regions than 12 other models in two test databases.