introduction
play

Introduction CptS 570 Machine Learning School of EECS Washington - PowerPoint PPT Presentation

Introduction CptS 570 Machine Learning School of EECS Washington State University What is Learning? Webster To gain knowledge or understanding of or skill in by study, instruction or experience To memorize Synonym: discover


  1. Introduction CptS 570 Machine Learning School of EECS Washington State University

  2. What is Learning? Webster � To gain knowledge or understanding of or skill in by study, instruction or � experience To memorize � Synonym: discover � To obtain knowledge of for the first time � May imply acquiring knowledge with little effort or conscious intention (as by � simply being told) or it may imply study and practice Knowledge � Knowing something with familiarity gained through experience or association � Facts or ideas acquired by study, investigation, observation, or experience � Deduction? (n!) � Knowledge representation? � Performance measure? �

  3. What is Machine Learning? � Herbert Simon, CMU � Any process by which a system improves its performance � Expert systems � Acquisition of explicit knowledge � Psychologists � Skill acquisition � Scientists � Theory formation, hypothesis formation and inductive inference � Tom Mitchell, CMU � A computer program that improves its performance at some task through experience

  4. Motivations � Automated knowledge engineering � Expertise is scarce � Codification of expertise is difficult � Expertise frequently consists of a set of test cases � Data from measurements, but no information or knowledge � Only one computer has to learn, then copy � Discover new knowledge � Understand human learning

  5. Applications � Speech recognition � Object recognition � Language learning � Autonomous navigation � Data mining � Intelligent agents � Cognitive modeling

  6. History � Exploration (1950s and 1960s) � Neurophysiological � Rosenblatt's perceptron � Biological � Simulated evolution � Psychological � Symbol processing systems � Statistical � Control and pattern recognition � Samuel's checkers program � Theoretical � Gold's identification in the limit � Minsky and Papert's criticism of the perceptron

  7. History � Development of practical algorithms (1970s) � Winston's ARCH � Learned concept of a blocks-world arch � Buchanan and Mitchell's Meta-Dendral � Learned mass-spectrometry prediction rules � Michalski's AQ11 � Learned soybean disease diagnosis rules � Quinlan's ID3 � Learned chess end-game rules � Fikes, Hart and Nilsson's MACROPS � Learned macro-operators in blocks-world planning � Lenat's AM � Discovered interesting mathematical concepts

  8. History � Explosion of research directions (1980s) � Learning theory � Symbolic learning algorithms � Connectionist (neural network) learning algorithms � Clustering and discovery � Explanation-based learning � Knowledge-guided inductive learning � Analogical and case-based reasoning � Genetic algorithms

  9. History � Maturity of the field (1990s) � Statistical comparisons of algorithms � Theoretical analyses of algorithms � Machine learning = Data mining (?) � Successful applications � Multi-relational learning � Ensemble and Kernel Methods

  10. Mitchell’s Book � Practical approach to study of machine learning � Methodology snapshot (good one for 1997)

Recommend


More recommend