Course Overview and Introduction CE-717 : Machine Learning Sharif University of Technology. M. Soleymani Fall 2014

Transcription

1 Course Overview and Introduction CE-717 : Machine Learning Sharif University of Technology M. Soleymani Fall 2014

2 Course Info Instructor: Mahdieh Soleymani Lectures: Sun-Tue (13:30-15) Website: TAs: Hassan Hafez Nooshin Maghsoudi Amin Sabzmakan Marzieh Gheisari 2

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

4 Marking Scheme Midterm Exam: 25% Final Exam: 35% Project: 10% Homeworks (written & programming) : 20% Mini-exams: 10% 4

5 Machine Learning (ML) and Artificial Intelligence (AI) ML appears first as a branch of AI ML is now also a preferred approach to other subareas of AI Perception (Computer Vision, Speech Recognition, ) Robotics Natural Language Processing ML is a strong driver in Computer Vision and NLP 5

6 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 6

7 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. 7

8 Some Learning Applications Face, speech, handwritten character recognition Document classification and ranking Self-customizing programs (e.g., recommender systems) Database mining (e.g., medical records) Market prediction (e.g., stock/house prices) Computational biology (e.g., annotation of biological sequences) Autonomous vehicles 8

9 Handwritten Digit Recognition

10 ML in Computer Science Why ML applications are growing? Improved machine learning algorithms Availability of data (Increased data capture, networking, etc) Demand for self-customization to user or environment Software too complex to write by hand 10

11 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 Other paradigms: semi-supervised learning, active learning, etc. 11

12 Experience (E) in ML We have different types of (getting) experience in different paradigms of ML methods (previous slide) 12

13 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 13

14 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 14

15 Supervised Learning: Example x 2 15? x 1 x 1 x 2 y

16 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/instances/samples Y column: Target/label Sample 2 Sample n-1 Sample n Evaluation 16 Test data?

17 Unsupervised Learning Given: Training set x i N i=1 Goal: find groups or structures in the data 17

18 Unsupervised Learning: Example x 2 Clustering 18 x 1

19 Sample Data in Unsupervised Learning Unsupervised Learning: x 1 x 2... x d Sample1 Columns: Features/attributes/dimensions Rows: Data/instances/samples Sample 2 Sample n-1 Sample n 19

20 Unsupervised Learning: Example Applications Clustering docs 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 20

21 Supervised Learning: Regression vs. Classification Supervised Learning Regression: predict a continuous target variable E.g., y [0,1] Classification: predict a discrete target variable E.g.,y {1,2,, C} A core objective of learning is to generalize from the experience. Generalization: ability of a learning algorithm to perform accurately on new, unseen examples after having experienced. 21

22 Regression: Example Housing price prediction 400 Price ($) in 1000 s Size in feet 2 Figure adopted from slides of Andrew Ng 22

23 Classification: Example Weight (Cat, Dog) 1(Dog) 0(Cat) weight weight 23

24 Main Steps of (Supervised) Learning Tasks Hypothesis class or model specification Which class of models (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 How do we ensure that the error on future data is minimized? (generalization) 24

25 Main Topics of the Course Supervised learning Regression Classification (we focus on this topic and introduce many classification methods) Model evaluation and selection Learning theory Ensemble learning Unsupervised learning Density estimation, unsupervised dimensionality reduction, and clustering Reinforcement learning Some advanced topics & applications 25