Course Overview and Introduction CE-717 : Machine Learning Sharif University of Technology. M. Soleymani Fall 2012

Transcription

1 Course Overview and Introduction CE-717 : Machine Learning Sharif University of Technology M. Soleymani Fall 2012

2 Course Info Instructor: Mahdieh Soleymani Lectures: Sun-Tue (13:30-15) Website: Office Hours: contact me to schedule an appointment. 2

3 Text Books Pattern Recognition and Machine Learning, C. Bishop, Springer, Machine Learning, T. Mitchell, MIT Press,1998. Additional readings: will be made available when appropriate. Other books: The elements of statistical learning, T. Hastie, R. Tibshirani, J. Friedman, Second Edition, Machine Learning: A Probabilistic Perspective, Kevin Murphy, MIT Press,

4 Marking Scheme Mid Term Exam: 25% Final Exam: 35% Project: 15% Homeworks (written & programming) : 15% Two quizzes: 10% 4

5 Main Topics of Course Regression Classification Learning theory Graphical models Unsupervised learning MDPs & Reinforcement learning Ensemble methods Active learning Semi-supervised learning Applications & some advanced topics 5

6 Machine Learning (ML) and Artificial Intelligence (AI) ML appears first as a field of AI ML is now also a preferred approach to other subareas of AI Perception (Vision, Speech Recognition, ) Robotics Natural Language Processing Reasoning 6

7 Learning Applications Face, speech, handwritten character recognition Autonomous vehicles Market prediction (e.g., stock/house prices) Self-customizing programs (recommender systems) Database mining (e.g., medical records) Document classification and ranking Annotation of biological sequences, molecules, or assays etc 7

8 Handwritten digit recognition 8

9 ML in Computer Science Why ML applications are growing? Improved machine learning algorithms Increased data capture, networking, new sensors Software too complex to write by hand Demand for self-customization to user, environment 9

10 ML Definition Tom Mitchell (1998): Well-posed learning problem A computer program is said to learn from experience E with respect to some task T and some performance measure P, if its performance on T, as measured by P, improves with experience E. Using the observed data to make better decisions Generalizing from the observed data 10

11 ML definition: example Consider an program that learns how to better filter spam according to s you do or do not mark as spam. T: Classifying s as spam or not spam. E: Watching you label s as spam or not spam. P: The number (or fraction) of s correctly classified as spam/not spam. 11

12 ML definition: example Consider an program that learns how to better filter spam according to s you do or do not mark as spam. T: Classifying s as spam or not spam. E: Watching you label s as spam or not spam. P: The number (or fraction) of s correctly classified as spam/not spam. 12

13 Paradigms of ML Supervised learning (regression, classification) predicting a target variable for which we get to see examples. Unsupervised learning revealing structure in the observed data Reinforcement learning partial (indirect) feedback, no explicit guidance Given rewards for a sequence of moves to learn a policy and utility functions Others: semi-supervised learning, active learning, recommender systems 13

14 Experience (E) in ML We have different types of (getting) experience in different paradigms of ML methods (previous slide) 14

15 An assumption Data are usually considered as vectors in a d dimensional space Now, we make this assumption for illustrative purpose We will see it is not necessary 15

16 Supervised learning x 2? x 1 16

17 Sample data in supervised learning Supervised Learning: right answers (targets) are known for training samples Columns: Features/attributes/dimensions Sample1 x 1 x 2... x d y (Target) Rows: Data/points/instances/examples/s amples Y column: Target/outcome/response/label Sample 2 Sample n-1 Sample n Evaluation 17 Test data?

18 Supervised learning Given: Training set labeled set of N input-output pairs D = x i, y i i=1 N Goal: learning a mapping from x to y 18

19 Unsupervised learning x 2 Clustering x 1 19

20 Sample data in unsupervised learning Unsupervised Learning: x 1 x 2... x d Sample1 Columns: Features/attributes/dimensions Rows: Data/points/instances/examples/s amples Sample 2 Sample n-1 Sample n 20

21 Unsupervised learning Given: Training set x i N i=1 Goal: find interesting patterns in the data 21

22 Unsupervised learning: examples Clustering documents based on their similarities Market segmentation: group customers into different market segments given a database of customer data. Finding semantic relations between ontological concepts in the molecular biology domain 22

23 Supervised Learning: Regression vs. Classification Supervised Learning Regression: predict a continuous target variable E.g., y (i) [0,1] Classification: predict a discrete target variable E.g.,y (i) *1,2,, C} 23

24 Regression: example Housing price prediction 400 Price ($) in 1000 s Size in feet 2 Figure adopted from slides of Andrew Ng 24

25 Classification: example Weight (Cat, Dog) 1(Dog) 0(Cat) weight weight 25

26 Main steps of (supervised) learning tasks Hypothesis class or model selection Which model (class of mappings) should we use for our data? Learning: find mapping f (from hypothesis class) based on the training set of examples Which notion of error should we use? (loss functions) Optimization of loss function to find mapping f Evaluation: how well f generalizes to yet unseen examples 26 How do we ensure that the error on future data is minimized? (generalization)