Prof Guenther Palm
Department of Neural Information Processing, Computer Science Faculty
University of Ulm, Germany

Title: Associative Memories for Technical Systems

Abstract:
The talk is about Neural Associative Memories (NAMs), which are very simple
neural networks that can be used for hetero- or auto-associative storage
and retrieval of information. After explaining what they are, I
will analyse their performance in terms of two major criteria:
retrieval time, and storage capacity. The latter is defined in information-
theoretical terms. I will comment on the various definitions of memory
capacity that some time are misused in the literature.