Admin Stuff 1 Artificial Intelligence 2 Machine Learning CSCI 4202 Spring 2006 Greg Grudic Greg Grudic Machine Learning 1 Location TR 2 00pm 3 15 pm ECCR 133 Instructor Greg Grudic Office ECOT 525 Office Hours Tuesday and Thursday 3 20 to 3 50 Tuesday and Thursday 11 30 to 12 00 And By Appointment Phone 303 492 4419 Email grudic cs colorado edu Course URL http www cs colorado edu grudic teaching CSCI4202 20 06 Greg Grudic Machine Learning 2 Admin Stuff 3 Admin Stuff 2 Homework 3 assignments each worth 10 Some coding will be required Most coding will be done in Matlab Course Textbook There isn t one but a good reference book is The Elements of Statistical Learning by Hastie Tibshirani Friedman Grading Homework Project Class participation Quizzes Matlab is available in department machines Project Pick by Feb 15 Application of machine learning for automating tasks that humans are good at and that cannot be easily programmed by hand human to robot skill transfer 30 50 10 10 OR if you have something in mind come see me You are encouraged to work in Groups Class participation This consists of showing up for class and asking questions Questions by email count as class participation Course workload outside of class About 4 hours per week Greg Grudic Machine Learning 3 Greg Grudic Machine Learning 4 1 Goal of the Course Where can ML be found A fundamental understanding of the basic concepts behind machine learning Marketing Profiling What does it mean for a machine to learn User interfaces Document characterization Bioinformatics Who should a company target for advertising Is passenger 57 likely to hijack the plane Making it easier to interact with a PC by anticipating what I am doing Why is machine learning important Searching the web for things of interest ML algorithms are at the heart of many modern computer applications Human genome project Which gene is responsible for the cancer that runs in my family Data mining Data doubles every year Dunham 2002 ML algorithms are used to make sense of this data Greg Grudic Machine Learning 5 Economics medical diagnosis robotics computer vision manufacturing inventory control elevator operation Greg Grudic ML is part of Artificial Intelligence What is AI Machine Learning My Personal View of AI The automation of activities that we associate with human thinking activities such as decision making problem solving learning Bellman 1978 The study of mental faculties through the use of computational models Charniak and McDermott 1985 The study of how to make computers do things at which at the moment people are better Rich and Knight 1991 The branch of computer science that is concerned with the automation of intelligent behavior Luger and Stubblefield 1993 I want to build a robot that will Greg Grudic Greg Grudic Machine Learning 6 7 Clean my house Cook when I don t want to Wash my clothes Cut my grass Fix my car or take it to be fixed i e do the things that I don t feel like doing Therefore AI is to me the science of building machines agents that act rationally with respect to a goal Machine Learning 8 2 What is a Rational Agent Elements of AI An agent is an entity that perceives and acts A rational agent is one that does the right thing Representation Reasoning The right thing that which is expected to maximize goal achievement accomplishing tasks that Greg doesn t feel like doing given the available information This is not a new idea Learning Aristotle Nicomachean Ethics Every art and every inquiry and similarly every action and pursuit is thought to aim at some good Greg Grudic Machine Learning 9 Machine Learning 10 Why Must Representation and Reasoning be Encompassed by Learning My Elements of AI Representation Greg Grudic How can I reason rationally about a world I know nothing about How can an gain knowledge about a world without sampling it and learning from those samples Fundamental lesson of AI learned in the 1980 s Reasoning It is not possible to hand code knowledge about anything but the most trivial problem domains Expert Systems largely failed because an expert e g doctor doesn t know how to formalize code what makes her an expert For Example I m an expert on chairs but I can t and no one can write a program that identifies chairs in an image Learning ML techniques can Greg Grudic Machine Learning 11 Greg Grudic Machine Learning 12 3 What is Machine Learning A Generic System x1 x2 System h1 h2 h K Tom Dietterich y1 y2 xN The goal of machine learning is to build computer systems that can adapt and learn from their experience yM Input Variables x x1 x2 xN What does this mean When are ML algorithms NOT needed Greg Grudic Machine Learning Hidden Variables h h1 h2 hK Output Variables y y1 y2 yK 13 Another Definition of Machine Learning Greg Grudic Machine Learning 14 When are ML algorithms NOT needed Machine Learning algorithms discover the relationships between the variables of a system input output and hidden from direct samples of the system When the relationships between all relevant system variables input output and hidden is adequately understood These algorithms originate form many fields This is NOT the case for almost any complex real system Statistics mathematics theoretical computer science physics neuroscience etc Greg Grudic Machine Learning 15 Greg Grudic Machine Learning 16 4 Class Project Human to Robot Skill Transfer The Robot Project Motivation Open Research problem in robotics Place a robot in an outdoor environment it has never been in and have it successfully move towards a GPS coordinate Humans can easily accomplish this task using only the robot s sensors vision and actuators wheels The Outdoor Environments This is a fundamental research problem with no known solution Currently the US government is putting hundreds of millions of dollars of basic research funding into this problem Greg Grudic Machine Learning 17 Greg Grudic Class Project Human to Robot Skill Transfer Use a human to teach the robot what it can traverse and what it cannot traverse These learning examples will be given to state of the art machine learning algorithms which will be used to build models of human behavior These models will then autonomously control the robot no human in the loop Greg Grudic Machine Learning Machine Learning 18 Class Project OR if you don t like robots Come up with at proposal requiring equal effort Key aspect of a project Successfully being able to use state of the art Machine Learning algorithms These are in general freely