Machine Learning CPSC 315 – Programming Studio Spring 2009 Project 2, Lecture 5.

Slides:



Advertisements
Similar presentations
Learning from Observations Chapter 18 Section 1 – 3.
Advertisements

Machine Learning: Intro and Supervised Classification
Adversarial Search We have experience in search where we assume that we are the only intelligent being and we have explicit control over the “world”. Lets.
CPSC 502, Lecture 15Slide 1 Introduction to Artificial Intelligence (AI) Computer Science cpsc502, Lecture 15 Nov, 1, 2011 Slide credit: C. Conati, S.
My name is Dustin Boswell and I will be presenting: Ensemble Methods in Machine Learning by Thomas G. Dietterich Oregon State University, Corvallis, Oregon.
Artificial Intelligence in Game Design Heuristics and Other Ideas in Board Games.
Learning from Observations Copyright, 1996 © Dale Carnegie & Associates, Inc. Chapter 18 Spring 2004.
Multiple Criteria for Evaluating Land Cover Classification Algorithms Summary of a paper by R.S. DeFries and Jonathan Cheung-Wai Chan April, 2000 Remote.
1 Chapter 10 Introduction to Machine Learning. 2 Chapter 10 Contents (1) l Training l Rote Learning l Concept Learning l Hypotheses l General to Specific.
Learning from Observations Copyright, 1996 © Dale Carnegie & Associates, Inc. Chapter 18 Fall 2005.
LEARNING FROM OBSERVATIONS Yılmaz KILIÇASLAN. Definition Learning takes place as the agent observes its interactions with the world and its own decision-making.
Learning from Observations Chapter 18 Section 1 – 4.
18 LEARNING FROM OBSERVATIONS
Learning From Observations
Learning from Observations Copyright, 1996 © Dale Carnegie & Associates, Inc. Chapter 18 Fall 2004.
Induction of Decision Trees
Learning from Observations Copyright, 1996 © Dale Carnegie & Associates, Inc. Chapter 18.
Evaluating Hypotheses
Three kinds of learning
LEARNING FROM OBSERVATIONS Yılmaz KILIÇASLAN. Definition Learning takes place as the agent observes its interactions with the world and its own decision-making.
Learning: Introduction and Overview
Reinforcement Learning Game playing: So far, we have told the agent the value of a given board position. How can agent learn which positions are important?
Hazırlayan NEURAL NETWORKS Radial Basis Function Networks I PROF. DR. YUSUF OYSAL.
Part I: Classification and Bayesian Learning
Chapter 5 Data mining : A Closer Look.
Radial Basis Function Networks
More Machine Learning Linear Regression Squared Error L1 and L2 Regularization Gradient Descent.
Minimax Trees: Utility Evaluation, Tree Evaluation, Pruning CPSC 315 – Programming Studio Spring 2008 Project 2, Lecture 2 Adapted from slides of Yoonsuck.
Machine Learning CPS4801. Research Day Keynote Speaker o Tuesday 9:30-11:00 STEM Lecture Hall (2 nd floor) o Meet-and-Greet 11:30 STEM 512 Faculty Presentation.
INTRODUCTION TO MACHINE LEARNING. $1,000,000 Machine Learning  Learn models from data  Three main types of learning :  Supervised learning  Unsupervised.
Learning CPSC 386 Artificial Intelligence Ellen Walker Hiram College.
Game Playing Chapter 5. Game playing §Search applied to a problem against an adversary l some actions are not under the control of the problem-solver.
Machine Learning1 Machine Learning: Summary Greg Grudic CSCI-4830.
Artificial Intelligence in Game Design Lecture 22: Heuristics and Other Ideas in Board Games.
Chapter 9 – Classification and Regression Trees
ENM 503 Lesson 1 – Methods and Models The why’s, how’s, and what’s of mathematical modeling A model is a representation in mathematical terms of some real.
Learning from observations
Learning from Observations Chapter 18 Through
CHAPTER 18 SECTION 1 – 3 Learning from Observations.
 2003, G.Tecuci, Learning Agents Laboratory 1 Learning Agents Laboratory Computer Science Department George Mason University Prof. Gheorghe Tecuci 5.
Learning from Observations Chapter 18 Section 1 – 3, 5-8 (presentation TBC)
Learning from Observations Chapter 18 Section 1 – 3.
CPS 270: Artificial Intelligence Machine learning Instructor: Vincent Conitzer.
Learning from observations
Chapter 11 Statistical Techniques. Data Warehouse and Data Mining Chapter 11 2 Chapter Objectives  Understand when linear regression is an appropriate.
Decision Trees Binary output – easily extendible to multiple output classes. Takes a set of attributes for a given situation or object and outputs a yes/no.
Goal of Learning Algorithms  The early learning algorithms were designed to find such an accurate fit to the data.  A classifier is said to be consistent.
Data Mining and Decision Support
Learning Kernel Classifiers 1. Introduction Summarized by In-Hee Lee.
Chapter 18 Section 1 – 3 Learning from Observations.
Machine Learning Lecture 1: Intro + Decision Trees Moshe Koppel Slides adapted from Tom Mitchell and from Dan Roth.
Learning From Observations Inductive Learning Decision Trees Ensembles.
SUPERVISED AND UNSUPERVISED LEARNING Presentation by Ege Saygıner CENG 784.
Anifuddin Azis LEARNING. Why is learning important? So far we have assumed we know how the world works Rules of queens puzzle Rules of chess Knowledge.
Decision Tree Learning CMPT 463. Reminders Homework 7 is due on Tuesday, May 10 Projects are due on Tuesday, May 10 o Moodle submission: readme.doc and.
Learning from Observations
Learning from Observations
Machine Learning Inductive Learning and Decision Trees
Introduce to machine learning
Presented By S.Yamuna AP/CSE
Classification and Prediction
Dr. Unnikrishnan P.C. Professor, EEE
Lecture 6: Introduction to Machine Learning
Learning from Observations
CS639: Data Management for Data Science
Lecture 14 Learning Inductive inference
Learning from Observations
Machine Learning: Decision Tree Learning
Instructor: Vincent Conitzer
Minimax Trees: Utility Evaluation, Tree Evaluation, Pruning
Presentation transcript:

Machine Learning CPSC 315 – Programming Studio Spring 2009 Project 2, Lecture 5

Forms of Learning Supervised Learns from examples which provide desired outputs for given inputs Unsupervised Learns patterns in input data when no specific output values are given Reinforcement Learns by an indication of correctness at end of some reasoning

Supervised Learning Must have training data including Inputs (features) to be considered in decision Outputs (correct decisions) for those inputs Inductive reasoning Given a collection of examples of function f, return a function h that approximates f Difficulty: many functions h may be possible Hope to pick function h that generalizes well Tradeoff between the complexity of the hypothesis and the degree of fit to the data Consider data modeling

Evaluating Supervised Learning Algorithms Collect a large set of examples (input/output pairs) Divide into two disjoint sets Training data Testing data Apply learning algorithm to training data, generating a hypothesis h Measure % of examples in the testing data that are successfully classified by h (or amount of error for continuously valued outputs) Repeat above steps for different sizes of training sets and different randomly selected training sets

Decision Trees Map features of situation to decision Example from a classification of unsafe acts:

Decision Trees Relation to rule-based reasoning Features of element used to classify element Features of situation used to select action Used as the basis for many “how to” books How to identify type of snake? Observable features of snake How to fix an automobile? Features related to problem and state of automobile If features are understandable, the decision tree can be used to explain decision

Learning Decision Trees Types of Decision Trees Learning a discrete-valued function is classification learning Learning a continuous-valued function is regression Assumption: use of Ockham’s razor will result in more general function Want the smallest decision tree, but that is not tractable Will be satisfied with smallish tree

Algorithm for Decision Tree Learning Basic idea Recursively select feature that splits data (most) unevenly No need to use all features Heuristic approach Compare features for their ability to meaningfully split data Feature-value = greatest difference in average output value(s) * size of smaller subset Avoids splitting out individuals too early

Unsupervised Learning Used to characterize/explain the key features of a set of data No notion of desired output Example: identifying fast-food vs. fine-dining restaurants when classes are not known ahead of time Techniques Clustering (k means, HAC) Self-Organizing Maps Gaussian Mixture Models More on this topic in Project 3

Reinforcement Learning Many large problems do not have desired outputs that can be used as training data Process Agent (system) performs a set of actions Agent occasionally receives a reward to indicate something went right or penalty to indicate something went wrong Agent has to learn relationship between the model of the situation, the chosen actions, and the rewards/penalties

Analogy to Animal Training We cannot tell our pets what is right and wrong in (preconditions, action) pairs Instead we reward good behavior (giving treats) and penalize bad behavior (spraying water or loud noise) Pet has to learn when and where what is appropriate Can result in incorrect interpretations (go in corner vs. go outside) Difficulty: what of the prior/recent actions caused the positive/negative outcome Clicker training for animals is meant to help this

Reinforcement Learning in Games Simplest reinforcements Winning or losing Requires lots of games/time to learn Other potential reinforcements Opponent’s action selection Did they minimize your goodness value Modify goodness function to better match their moves Potential to learn an individual’s values/strategy Predicted goodness value vs. observed goodness value This can be used in small (a few moves) or large (a game) time scales Similar to person reflecting on when things went wrong Need to be careful in implementation or else goodness function will return a constant (thus being totally consistent)

Modifying a Goodness Function Consider the game of chess Presume goodness function has three linear components BoardControl the difference between the number of board positions that Player1 and Player2 can get a piece to in one move Threatened the difference between the number of opponents pieces threatened (can be taken in one move) between Player1 and Player2 Pieces the difference in the sum of the values of pieces left for Player1 and Player2 where Queen = 10, Rook = 6, Bishop = 3, Knight = 3, Pawn = 1

Modifying a Goodness Function G(s) = a*BoardControl + b*Pieces + c*Threatened Modify coefficients to learn appropriate weighting of terms Quantity of overall modification should relate to difference between predicted goodness and observed goodness Direction of modification to each linear component should be related to whether they are consistent with or disagree with outcome Could modify coefficients using fixed values (e.g. +/-.1) or with values a function of their effect on overall G for the state being considered In theory, such a computer player could recognize that BoardControl is more important early in a game, Pieces is more important mid-game, and Threatened is more important for the end game.