Unformatted text preview:

1ConclusionsLarry HolderCSE 6363 – Machine LearningComputer Science and EngineeringUniversity of Texas at Arlington2Outline Overview of machine learning Machine learning and data mining Fundamental research issues Grand challenge problems3Overview of Machine Learning Supervised learning Evaluation of learning methods Learning theory Unsupervised learning Other learning methods Applications Related fields4Supervised Learning Traditional methods Version space Candidate elimination algorithm Decision tree induction Neural networks Bayesian learning Instance-based learning5Supervised Learning Advanced methods Kernel methods Support vector machines Ensembles Bagging Boosting Learning rule sets Relational learning Inductive logic programming (ILP) Graph-based learning6Evaluation of Learning Methods True error vs. sample error Bounding true error Comparison of hypotheses Comparison of learners Significance testing ROC curves7Learning Theory Bayes optimal learning Sample complexity PAC learning framework VC dimension Mistake bound framework8Unsupervised Learning Pattern discovery Clustering Grammar (language) learning Self-organizing maps (SOMs) EM algorithm9Other Learning Methods Genetic algorithms Analytical learning Reinforcement learning Integrated learning10Applications Classification and prediction Chemical properties Biometrics Object recognition Organizational and behavioral patterns Skill acquisition Robot navigation Control and optimization Heuristic search11Related Fields Statistics Pattern recognition Control theory Cognitive science Psychology Neurophysiology12Machine Learning andData Mining Knowledge Discovery and Data mining (KDD) process13Machine Learning andData Mining “Data mining” now synonymous with “KDD process” Data mining also emphasizes Database (particularly large DB) issues Producing many relevant patterns, rather than only the best14Fundamental Research Issues General learning methods Limits of general methods Theory and principles guiding development of domain-specific learning algorithms Non-propositional learning Learning in dynamic environments Incorporation of domain-specific background knowledge Ethical responsibility and privacy15Grand Challenge Problems Mitchell et al., “Machine Learning,” Annual Review of Computer Science, Volume 4, 1990. Characterize goals and potential capabilities Motivate the need for continued research16Grand Challenge Problems Learning household robot to assist the handicapped Ability to operate in complex, unknown and changing environments Path planning, obstacle avoidance, low-level perception, and manipulation Recognize and manipulate specific objects Model changing environment including expectations Strategies for specific problem solving Input from observation, advice, and experimentation17Grand Challenge Problems Learning assembly robot for flexible manufacturing Perceive and manipulate new parts Assess physical limitations of materials Generalization of specific training actions18Grand Challenge Problems Learning spoken-dialog system for advising on equipment repair Given schematics and behavior of components Assist and apprentice to humans performing the task Human-machine collaborative problem solving Recognize new speakers, accents, dialects, known speakers under new conditions New vocabulary and grammar (including colloquialisms) Model of user New troubleshooting strategies19Grand Challenge Problems System that learns by reading and practicing Read a chapter of physics or calculus book and answer questions at the end of the chapter Natural language and problem-solving Better models of textbook learning Read instruction manuals and troubleshoot20Grand Challenge Problems Self-compiling expert systems: learning expert system for engineering design Given physics, design requirements and constraints, manufacturing and assembly constraints Reduce design problems to routine design rules21Grand Challenge Problems Automated discovery of important regularities in scientific databases DNA sequences, protein folding, astronomical data Inclusion of domain knowledge22Ultimate Grand Challenge What ever happened to Voyager?“Star Trek - The Motion Picture” Paramount Pictures,


View Full Document

WSU CSE 6363 - Study Notes

Download Study Notes
Our administrator received your request to download this document. We will send you the file to your email shortly.
Loading Unlocking...
Login

Join to view Study Notes and access 3M+ class-specific study document.

or
We will never post anything without your permission.
Don't have an account?
Sign Up

Join to view Study Notes 2 2 and access 3M+ class-specific study document.

or

By creating an account you agree to our Privacy Policy and Terms Of Use

Already a member?