Neural computation methods and applications: Summary talk of the AI session

The sixty papers selected for oral presentation in the AI session of the workshop were mostly application oriented, with high energy physics taking central stage. The most popular methodology was that of neural networks. In this summary talk I point out some general trends of the research work presented at the workshop, and concentrate on papers that I found especially interesting. I add sections that deal with recent developments in neural computation techniques that may be of relevance to applications in the future. The latter include PCA and ICA analyses, averaging over estimators and models of spiking neurons. Touching on some general questions, I indicate the advantages of the neural computation approach.