Lecture 0: Machine Learning

Transcription

1 Lecture 0: Machine Learning Tuo Zhao Schools of ISYE and CSE, Georgia Tech 2017 Fall

2 Questions Course Logistics Why Machine Learning? What is a well-defined learning problem? What questions should we ask about Machine Learning? Tuo Zhao Lecture 0: Machine Learning 2/22

3 Machine Learning is Interdisciplinary Tuo Zhao Lecture 0: Machine Learning 3/22

4 Pre-requisites Math: Calculus and Linear Algebra Probability and Statistics Basic Optimization Coding: MATLAB for coding HW (No Exception) Plus: Generalized Linear Models Convex Optimization Tuo Zhao Lecture 0: Machine Learning 4/22

5 Course Logistics Teaching Assistants: Shaojun Ma: Ph.D. Student in CEE Yujia Xie: Ph.D. Student in CSE Minshuo Chen: Ph.D. Student in ISYE Haoming Jiang: Ph.D. Student in ISYE Zhehui Chen: Ph.D. Student in ISYE TBD See Syllabus, Lecture Slides Homework Assignments Tuo Zhao Lecture 0: Machine Learning 5/22

6 Highlights of Course Logistics Working Load: Background Knowledge Test: 6% 4 Written HW: 24% 3 Coding HW: 18% Exam-1: 26% Exam-2: 26% See Relase Important Announcements! MUST Register! You can post anonymously (to other students, but not me) Tuo Zhao Lecture 0: Machine Learning 6/22

7 Distance Learning Working Load: 5 Written HW: 60% 4 Coding HW: 40% No Background Knowledge Test No Mid-term Exam Late Homework Policy for All Students: No Late Homework Accepted! Always due at noon on Friday Tuo Zhao Lecture 0: Machine Learning 7/22

8 Knowledge Background Test Statistics Top 10%: 30/40 Top 25%: 27/40 Top 50%: 20/40 Top 75%: 14/40 Maximum: 38 Suggestions Go through the review materials carefully! Tuo Zhao Lecture 0: Machine Learning 8/22

9 Remarks Office hours for asking questions Sep. 19 A more difficult make-up exam (but will be curved accordingly) No time for answering questions after class You need to debug by yourself Honor Code Tuo Zhao Lecture 0: Machine Learning 9/22

10 What to Cover? Methodology and Algorithms of Machine Learning. Some Theory for Ph.D. Students. Some homework problems will be for Ph.D. students ONLY. Different letter grades for each section. Not About Introduction to Machine Learning Not About How to Apply Machine Learning to Your Domain. Not About How to Use Software to Do Machine Learning Tuo Zhao Lecture 0: Machine Learning 10/22

11 Alternative Easier: CS 4641 Machine Learning Signal Processing: ECE 6254: Statistical Machine Learning Learning Theory: CS 7545 Machine Learning Theory More Foundation: CS 8803 Mathematical Foundations of Machine Learning Applications to Specific Domains: Computer Vision, Natural Language Processing, etc. Tuo Zhao Lecture 0: Machine Learning 11/22

12 Why Machine Learning? Recent progress in algorithms and theories Growing flood of massive data Computational power is available Budding industry Tuo Zhao Lecture 0: Machine Learning 12/22

13 Why Machine Learning? Three Niches for Machine Learning: Data mining: using historical data to improve decisions medical records medical knowledge Software applications we can t program by hand autonomous driving speech recognition Self customizing programs Newsreader that learns user interests Tuo Zhao Lecture 0: Machine Learning 13/22

14 What is the Learning Problem? Learning: Improving with experience at some task Improve over task T with respect to performance measure P based on experience E Example: Learn to play checkers T : Play checkers P : % of games won in world tournament E: opportunity to play against self Tuo Zhao Lecture 0: Machine Learning 14/22

15 ISYE6740/CSE6740/CS7641: Computational Data Analysis/Machine Learning Learn to to Predict Emergent C-sections Learning topredict Predict Emergency C-Sections Learning Emergency C-Sections Data: [Sims et al., 2000] [Sims et al., 2000] 9714 patient records, 9714 patient records, each with 215 features each with 215 features One of 18 learned rules: Tuo Zhao Lecture 0: Machine Learning 15/22

16 Learn to Detect Objects in Images Tuo Zhao Lecture 0: Machine Learning 16/22

17 Learn to Classify Documents Tuo Zhao Lecture 0: Machine Learning 17/22

18 Learn to Drive Autonomously Tuo Zhao Lecture 0: Machine Learning 18/22

19 Learn to Recognize Speech Tuo Zhao Lecture 0: Machine Learning 19/22

20 Learn to Translate Languages Tuo Zhao Lecture 0: Machine Learning 20/22

21 Learn to Play Computer Games Tuo Zhao Lecture 0: Machine Learning 21/22

22 Next 3 Lectures Linear Algebra Review (2 Lectures) Probability Review (1 Lecture) Tuo Zhao Lecture 0: Machine Learning 22/22