Comprehensible Descriptions for Fast Processing of Image Data

The proposed effort will apply advanced methods of hybrid learning to problems of detection and identification of sensory events (e.g., boost phase evens). Our approach aims at building a vision system capable of hybrid learning, in which symbolic (rule-based) and subsymbolic (neural network) strategies are integrated to achieve high efficiency and accuracy both in learning human comprehensible sensory descriptions and in employing them for recognition. The proposed approach has several advantages: it can be easily modified and applied to new problems (due to learning) it has fast recognition rates (due to parallel architecture), and its recognition algorithm is easy to understand by a human operator (due to the underlying symbolic knowledge representation).