"The machine learning professor who mapped the field's five warring tribes — and bet they're all pointing at the same answer."
Pedro Domingos is a Professor Emeritus of computer science and engineering at the University of Washington, where his research spans machine learning, data mining, and artificial intelligence. He is best known for developing Markov logic networks, a framework combining logical and probabilistic reasoning, and for winning the SIGKDD Innovation Award, the highest honor in data science. In The Master Algorithm, Domingos steps back from his own research to map the entire field of machine learning into five competing schools of thought, arguing that a single unifying algorithm connecting all five is both possible and closer than most researchers assume. His cross-disciplinary career — working fluently across symbolic AI, neural networks, and Bayesian methods — gives the book a rare authority few single-tribe researchers could match.
Our in-depth summaries and reviews of his work
Pedro Domingos maps machine learning's five competing tribes — symbolists, connectionists, evolutionaries, Bayesians, and analogizers — and argues they're all pointing at…
Read Summary →Domingos breaks all machine learning down to representation, evaluation, and optimization — the same three ingredients recombined differently by each tribe.
Symbolists, connectionists, evolutionaries, Bayesians, and analogizers each solve learning differently, but Domingos argues they're converging on the same target.
A model that memorizes its training data instead of the pattern behind it will fail on anything new — every tribe fights this same problem.
"Every algorithm has an Achilles heel."— The Master Algorithm
"The five tribes of machine learning are not really in competition. They're pieces of a puzzle."— The Master Algorithm
"To learn is to forget the details and remember what matters."— The Master Algorithm
Pedro Domingos maps machine learning's five competing tribes — symbolists, connectionists, evolutionaries, Bayesians, and analogizers — and argues they're all pointing at one unifying master algorithm.
