Pedro Domingos University of Washington
Traditional Programming Machine Learning Computer Data Algorithm Output Computer Data Output Algorithm
Traditional Programming Machine Learning Computer Data Algorithm Output Master Algorithm Data Output Algorithm
TribeOriginsMaster Algorithm SymbolistsLogic, philosophyInverse deduction ConnectionistsNeuroscienceBackpropagation EvolutionariesEvolutionary biologyGenetic programming BayesiansStatisticsProbabilistic inference AnalogizersPsychologyKernel machines
Tom MitchellSteve MuggletonRoss Quinlan
AdditionSubtraction ――― = ? ―― 2 + ? ――― = 4 ――
Deduction Socrates is human + Humans are mortal. ――――――――――― = ? Induction Socrates is human + ? ――――――――――― = Socrates is mortal ――――――――――
Yann LeCunGeoff HintonYoshua Bengio
John Koza John HollandHod Lipson
David HeckermanJudea PearlMichael Jordan
Peter HartVladimir VapnikDouglas Hofstadter
TribeProblemSolution SymbolistsKnowledge compositionInverse deduction ConnectionistsCredit assignmentBackpropagation EvolutionariesStructure discoveryGenetic programming BayesiansUncertaintyProbabilistic inference AnalogizersSimilarityKernel machines
TribeProblemSolution SymbolistsKnowledge compositionInverse deduction ConnectionistsCredit assignmentBackpropagation EvolutionariesStructure discoveryGenetic programming BayesiansUncertaintyProbabilistic inference AnalogizersSimilarityKernel machines But what we really need is a single algorithm that solves all five!
Representation Probabilistic logic (e.g., Markov logic networks) Weighted formulas → Distribution over states Evaluation Posterior probability User-defined objective function Optimization Formula discovery: Genetic programming Weight learning: Backpropagation
Much remains to be done... We need your ideas
Home Robots Cancer Cures360 o Recommenders World Wide Brains