Perceptrons: An Introduction to Computational Geometry

Perceptrons: An Introduction to Computational Geometry is a significant work by Marvin Minsky and Seymour Papert that critically examined the capabilities and limitations of perceptrons, a type of artificial neural network.12

Perceptrons, first investigated by Frank Rosenblatt in 1957, were early models in the field of artificial intelligence (AI) that aimed to simulate how the human brain learns and processes information through interconnected nodes, or neurons. Rosenblatt's work, which he termed "connectionism," focused on how these connections are formed and modified during learning. He also developed a method for training multi-layer networks, known as "back-propagating error correction".13

Minsky and Papert's book, however, highlighted that simple perceptrons were incapable of solving certain complex problems, such as pattern recognition tasks that required more than a single layer of processing. This analysis, published in 1969, significantly influenced the direction of AI research, leading to a period of reduced funding and interest in neural networks, often referred to as the "AI winter". Despite these critiques, the foundational concepts explored in Perceptrons remain relevant to the study of computational geometry and the development of modern AI systems.21