In this groundbreaking work, computer scientist Leslie G. Valiant details a promising new computational approach to studying the intricate workings of the human brain. Focusing on the brain's enigmatic ability to access a massive store of accumulated information very quickly during reasoning processes, the author asks how such feats are possible given the extreme constraints imposed by the brain's finite number of neurons, their limited speed of communication, and their restricted interconnectivity. Valiant proposes a "neuroidal model" that serves as a vehicle to explore these fascinating questions.
While embracing the now classic theories of McCulloch and Pitts, the neuroidal model also accommodates state information in the neurons, more flexible timing mechanisms, a variety of assumptions about interconnectivity, and the possibility that different areas perform different functions. Programmable so that a wide range of algorithmic theories can be described and evaluated, the model provides a concrete computational language and a unified framework in which diverse cognitive phenomena--such as memory, learning, and reasoning--can be systematically and concurrently analyzed.
Included in this volume is a new preface that highlights some remarkable points of agreement between the neuroidal model and findings in neurobiology made since that model's original publication. Experiments have produced strong evidence for the theory's predictions about the existence of strong synapses in cortex and about the use of precise timing mechanisms within and between neurons. The theory also provides a quantitative explanation of how randomly placed neurons can be harnessed as resources for general purpose learning and memory--and is therefore synergistic with the striking recent discovery of neurogenesis in cortex.
Requiring no specialized knowledge, Circuits of the Mind, masterfully offers an exciting new approach to brain science for students and researchers in computer science, neurobiology, neuroscience, artificial intelligence, and cognitive science.
"The book is written in a clear style, with a sufficient number of figures illustrating the algorithms. . .This new insight into complex problems of the brain, as well as the proposed methodology, makes the book highly readable and interesting." --Computing Reviews
"The author shows that the proposed neuroidal model supports the cognitive activities he identifies. It provides a good structure to explore the functions of the mind still further." -- IEEE Spectrum
"Although there are many books today dealing with a simple neuronal model based on the weighted sum principle, this one rises above these others in providing an explanation of cognitive functions." --Choice
1: The Approach
2: Biological Constraints
3: Computational Laws
4: Cognitive Functions
5: The Neuroidal Model
6: Knowledge Representation
7: Unsupervised Memorization
8: Supervised Memorization
9: Supervised Inductive Learning
10: Correlational Learning
11: Objects and Relational Expressions
12: Systems Questions
14: More Detailed Neural Models
Index of Notation
Number Of Pages: 254
Published: 1st December 2000
Publisher: Oxford University Press Inc
Country of Publication: US
Dimensions (cm): 23.39 x 15.6
Weight (kg): 0.43