Introduction to Machine Learning

Introduction to Machine Learning Fall 2014 The University of Iowa Tianbao Yang

Introduction • Instructor: Tianbao Yang • Office hour: • MWF: 1:20 pm – 2:20 pm or by appointment • Textbooks • Pattern Recognition and Machine Learning (C. M. Bishop) • Optional: The Elements of Statistical Learning (T. Hastie et al.) • Some subjects are from papers • TA: Shiyao Wang

Requirements • 6 homework assignments (every two weeks): 40% • written + programming • Course project: 50% • Topics: related to machine learning, suggestions will be given prior, your own choice • Collaboration: encourage 1~2 people • Proposal (10%) + Presentation (10%) + Report (30%) • No exam: yeah • Participation: 10%

Prerequisites • Linear Algebra • vector, matrix, norm, etc • Probability • probability, distribution, expectation, etc • Optimization • function, convex, gradient, etc • Review will be given if necessary

Course Content • Introduction • Linear Regression • Linear Classification • Kernel Methods • Generative and Discriminative Models • Mixture Models and EM • Regularization and Sparse Learning • Recent Advances in BIG DATA

Expectation • You are expected to know/master • what is machine learning • how to learn a regression model • how to learn a classification model • how to learn a clustering model • what are the recent advances in machine learning • what is BIG DATA and why BIG DATA

Today’s Topics • What is Machine Learning? • definition • procedure • categorization • history • Real World Applications

What is Machine Learning • Arthur Samuel (1959) • “Field of study that gives computers the ability to learn without being explicitly programmed” Computer Checkers Checkers http://en.wikipedia.org/wiki/Machine_learning

What Samuel did? “Samuel developed a scoring function based on the position of the board at any given time. This function tried to measure the chance of winning for each side at the given position. It took into account such things as the number of pieces on each side, the number of kings, and the proximity of pieces to being “kinged” ”-- wiki

What is Machine Learning • Tom M. Mitchell (CMU) • A computer program is said to learn from experience E with respect to some task T and performance measure P, if it • improves performance P at task T with experience E • Example: Computer Checkers • T (task): play checkers • P (performance): probability to win • E (experience): play the game with itself many times

Machine Learning and Data Mining • Differences • Different Goals • machine learning: prediction • data mining: discovery of unknown knowledge • e.g., Is the customer male or female based on what he/she purchased? (machine learning) • customers who buy diapers also tend to buy beer (data mining) • Similarities • both are data-driven • data mining uses many machine learning methods, machine learning also employ data mining methods

Procedures of Machine Learning • A Three-step view of Machine Learning • data collection (and pre-processing) • model building (and analysis) • optimization Data Optimization Model

Data • Training Data: a set of instances (examples) with • inputs (features, attributes) • quantitative (continuous) • qualitative (discrete) • output (label) • qualitative  classification • quantitative  regression • Typical task: • predict output based on inputs Training Data Learning Algorithm output inputs Optimal Model

Examples: Email Spam (Classification) • Email Spam • Training Data: a set of emails and their labels • input features: occurrence of certain words (in an email) • output label: spam or not spam or not? 1 1 0 1 0 0 1 0 0 1 0

Examples: House Price (Regression) • Predict House Price • Data: a set of houses with features and price • input features: size, year, etc • output: house price

Data • In math • input features: d-dimensional vector • output label:

Model • Model is a way to describe the relation between input features and output label • regression, classification • ranking, clustering, etc • in math:

Optimization • Search for the best model parameters • in the space of all parameters • subject to certain criterion (e.g., minimize the classification error on the given data, etc)

Categorization of Machine Learning • Categorization of Machine Learning Algorithms • based on tasks • regression (e.g. predict house price) • classification (e.g. predict spam) • clustering, ranking, etc • based on availability of labels • yes: supervised learning (classification, regression) • no: unsupervised learning (clustering) • partially: semi-supervised learning (classification, regression) • based on methodologies • discriminative • generative

History of Machine Learning Deep Learning scale 109 Logistic regression 106 Support vector machine 103 decision tree K nearest neighbor neural network 10 time 1960 1970 1980 1990 2000 2010

Earlier Days • Two Layer Neural Networks (1960 – 1970) • Perceptron (Frank Rosenblatt)

1980 – 2000 • Multiple layers Neural Network • Several layers • Application: digit recognition

1980 – 2000 color • Decision tree • Rule based model • Easy to understand • Application: commercial systems (credit risk analysis) yellow red size size big small big small Grape-fruit lemon Apple cherry

Since Mid-90’s • Support Vector Machine • Vladimir N. Vapnik • widely studied • Logistic Regression • Application: internet

Since 2010 • BIG DATA Analytics • Deep Learning • Large-scale Optimization

Regression • Predict the examination score of students • study the effectiveness of schools • Training Data • student-dependent features: • hours of study • gender, ethnic group • school-dependent features: • percentage of students eligible for free school meals • school gender, school denomination

Classification • Handwritten Recognition • Postal address recognition • more than 95% of handwritten mail is sorted automatically

Ranking • Google Search • given a query, find relevant webpages and rank in order

Recommendation • Netflix movie recommendation: predict users’ ratings • given users’ information, watching history, etc, recommend most-likely to watch movies • Netflix Prize: $1,000,000, 2006 – 2009

Speech Recognition • SIRI

Classification of Galaxy • understanding the universe

More Practical Problems • Kaggle Competition • predict which shoppers will become repeat buyers. Award:$30,000 • predict which ads contain illicit content. Award: $25,000 • predict click-through rates on display ads. Award: $16,000 • etc

Some Real Demos demo

Introduction to Machine Learning