Unsupervised training of deep generative models containing latent variables and performing inference remains a challenging problem for complex, high dimensional distributions. One basic approach to this problem is the so called Helmholtz machine and it involves training an auxiliary model that helps to perform approximate inference jointly with the generative model which is to be fitted to the ...

We present a computational model for visual processing where attentional focus emerges fundamental mechanisms inherent to human vision. Through detailed analysis of activation development in the network we demonstrate how normal interaction between top-down and bottom-up processing and intrinsic mutual competition within processing units can give rise to attentional focus. The model includes bo...

In our previous work (e.g., 4], 5], 6], 7]), we have formalized the story understanding process based on scripts and plans with stepwise default theories. While those theories ooer nal results for understanding a speciic story, they do not provide the history of changes of partial states of any objects the story may concern. Moreover, the causal models for missing events are incomplete in scrip...

For the past ten years I have been involved in industrial research on applications of agent architectures to manufacturing, and am presently techical advisor to an industrial consortium that is fielding pilot tests of shop-floor agents. The specific problem addressed by most of our work is the management of material flow and resource utilization on the factory floor. Classically, these problems...

The impact of induced technological change (ITC) in energy/climate models on the timing of optimal CO2-abatement depends on whether R&D or learning-by-doing (LBD) is the driving force. Bottom-up energy system models employing LBD suggest strong increases in optimal early abatement. In this paper we extend an existing top-down model supporting this view according to the notion that socio-economi...

By assuming network trafic to be independent fLom each other, the analysis of network pevformance can be simplified However, the real traflc sources may have some correlation which makes their behavior tends to converge or diverge. This phenomenon has tremendous influence on the congestion control. In this paper, we explain the possible reasons of the correlated behavior, namely, top-down and c...

Theories of visual attention commonly posit that early parallel processes extract conspicuous features such as color contrast and motion from the visual field. These features are then combined into a saliency map, and attention is directed to the most salient regions first. Top-down attentional control is achieved by modulating the contribution of different feature types to the saliency map. A ...

We propose a fully automatic framework to detect and extract arbitrary human motion volumes from real-world videos collected from YouTube. Our system is composed of two stages. A person detector is first applied to provide crude information about the possible locations of humans. Then a constrained clustering algorithm groups the detections and rejects false positives based on the appearance si...

This paper discusses the decidability of node query preservation problems for XML document transformations. We assume a transformation given by a deterministic linear top-down data tree transducer (abbreviated as DLTV ) and an n-ary query based on runs of a tree automaton. We say that a DLTV Tr strongly preserves a query Q if there is a query Q′ such that for every document t, the answer set of...

In some classification problems, apart from a good model, we might be interested in obtaining succinct explanations for particular classes. Our goal is to provide simpler classification models for these classes without a significant accuracy loss. In this paper, we propose some modifications to the splitting criteria and the pruning heuristics used by standard top-down decision tree induction a...