CS 445/545 Machine Learning Winter, 2012

1. CS 445/545 Machine LearningWinter, 2012 • Course overview: • Instructor • Melanie Mitchell • Textbook • Machine Learning: An Algorithmic Approach by Stephen Marsland

2. Homework • Bi-weekly assignments, programming or using existing code. • Do on your own; no teamwork, but you can discuss general ideas with classmates • Quizzes • Short quiz (20 minutes) every Wednesday (starting 1/25) to test basic understanding of the material presented in class and in the readings. • If you must miss a quiz, see me to schedule a make-up quiz. • Final exam • Take-home open-book exam.

3. Grading • Homework: 50% • In-class quizzes: 30% • Final: 20% • Mailing list: MachineLearning2012@cs.pdx.edu (please give your e-mail address on the sign-up sheet)

4. Academic Integrity • You can discuss concepts with other students, but all work you hand in must be your own. • You may not copy any text from other students, papers, the internet, or any other source, without clearly referencing it. • See syllabus for URL with PSU policy.

5. Late homework policy: • If needed, request an extension before the assignment is due. Otherwise, 5% of the assignment grade will be subtracted for each day the homework is late.

6. Use of laptops, phones, etc. in class: • Please don’t, unless you are using it to take notes or view class slides.

7. What is machine learning? • Textbook definitions of “machine learning”: • Detecting patterns and regularities with a good andgeneralizable approximation (“model” or “hypothesis”) • Execution of a computer program to optimize the parameters of the model using training data or past experience.

8. Training Examples: Class 1 Training Examples: Class 2 Test example: Class = ?

9. Training Examples: Class 1 Training Examples: Class 2 Test example: Class = ?

10. Training Examples: Class 1 Training Examples: Class 2 Test example: Class = ?

11. Training Examples: Class 1 Training Examples: Class 2 Test example: Class = ?

12. Training Examples: Class 1 Training Examples: Class 2 Test example: Class = ?

13. Training Examples: Class 1 Training Examples: Class 2 Test example: Class = ?

14. Any observations from these machine learning examples?

15. Types of machine learning tasks • Classification • Output is one of a number of classes (e.g., ‘A’) • Regression • Output is a real value (e.g., ‘$35/share”)

16. Types of Machine Learning Methods • Supervised • provide explicit training examples with correct answers • e.g. neural networks with back-propagation • Unsupervised • no feedback information is provided • e.g., unsupervised clustering based on similarity

17. “Semi-supervised” • feedback information is provided, but is not detailed • examples: • e.g., reinforcement learning: reinforcement single is single-valued assessment of current state

18. Relation between “artificial intelligence” and “machine learning”?

19. Key Ingredients for Any Machine Learning Method • Features (or “attributes”) • Underlying representation for “hypothesis”, “model”, or “target function”: • Hypothesis space:

20. Learning method: • Data: Divide into three parts. • Training data • Used to train the model • Validation data • Used to select model complexity, to determine when to stop training, or to alter training method • Test data • Used to evaluate trained model • Evaluation method

21. Demonstration • What is: • Representation: • Features: • Hypothesis space: • Learning method: • Data: • Evaluation method: • Notion of “confusion matrix”

22. Assumption of all ML methods: Inductive learning hypothesis: Any hypothesis that approximates target concept well over sufficiently large set of training examples will also approximate the concept well over other examples outside of the training set. Difference between “induction” and “deduction”?

23. What questions does ML ask? • Which is the best method for a given learning problem? • What can be proved about convergence, performance? • How much training data or feedback is needed for good generalization performance?

24. How to select/design training examples or feedback? • How can prior knowledge be used to guide learning? • How can meta-learning be used? (I.e., learner automatically modifies its representation, selection of training examples, etc.)

25. Syllabus

26. Homework 1