Neural network fundamentals of the hippocampus

The field of artificial intelligence explores the nature of learning and memory, two defining aspects of intelligence. Several strategies for approximating or simulating these facets are well known in AI, including the examination of biological neural circuits. In particular, analysis of neural network properties in the brain may provide insight into the mechanisms and algorithms that ultimately give rise to intelligence. The brain structure known as the hippocampus is often associated with tasks requiring learning and memory. Its well-documented features lend itself to computational analysis.; Yet much about the hippocampus remains poorly understood. Findings from several lines of study that improve our fundamental understanding of the anatomical and physiological bases of computation within the hippocampus will be presented. In Chapter 2, results clarifying the pattern of connectivity in field CAI are discussed along with their implications for neural networks.; Second, in Chapters 3 and 4, findings from hippocampal slice preparations exhibiting the electrical activity known as sharp waves (SPWs) are presented. The discovery of this EEG waveform in hippocampal field CA3, heretofore seen only in vivo, paves the way for systematic analyses of its origin and functional significance. In a follow up series of experiments, pharmacological tests showed that SPWs are mediated via AMPA and NMDA receptors and are insensitive to atropine, indicating a nondependence on muscarinic receptors. Identification of receptor types provides clues critical for deducing CA3 network properties and possible roles for sharp waves in memory formation.; In Chapter 5, evidence of cholinergic plasticity in the hippocampus are discussed. Strong, transient activation of muscarinic receptors in CA3 by washin of either carbachol or physostigmine resulted in elevated levels of rhythmic activity which persisted stably for hours after washout. These lasting rhythms were permanently abolished by infusion of atropine, but only temporarily by serotonin. Since oscillatory activity in the brain is widely believed to synchronize parallel efforts of distributed computation and assist in the induction of memory, this preliminary finding of sustained, rhythmic generation raises many network issues.; Finally, efforts to integrate these results into an overall picture of information processing in fields CA3 and CA1 will be discussed.