• Stanford CS229 Machine Learning Course Website
• Stanford Machine Learning Course Videos by Prof. Andrew Ang
• Stanford CS229 Course Handouts
• Machine learning projects done by Stanford CS229 students
• Octave software similar to MATLAB but free
• Univ  of British Columbia CPSC340 Machine Learning & Data Mining Course
• The Elements of Statistical Learning Book (PDF)
• AISpace machine learning tools online
• JavaBayes: App from CMU

Handwriting recognition, flying helicopter, checks, US Mail sorting, credit card fraud detection, classifying human genome, etc.

Field of study that gives computers the ability to learn without being explicitly programmed (Arthur Samuel, 1959 who wrote Checkers program that played 1000s of games with itself and learnt to recognize board positions that are good/bad and eventually played better than Arthur Samuel)

Recent Definition: A computer program is said to learn from experience E w.r.to some task T and some performance measure P, it its performance on T, as measured by P, improveds with experience E.

Providing the algorithm the "right answers" for a number of cases.

In classification problems, the right answers are discrete values (e.g., Is tumor malignant?)

How and why learning algorithms work? We can prove that a learning algorithm will work with X% accuracy and what conditions. Do we have enough data or need more? How to take the tools of machine learning and apply to real problems.

No right answers given -- the program finds structure in the data such as clusters. Applcations include image processing to group pixels; group gene data; market segmentation; social network analysis; astronomical data analysis. An interesting application (using independent component analysis) is seperating the voices of individuals at a cocktail party using multiple microphones.

Specify what needs to be done (good dog, bad dog) and the algorithm maximizes the reward.

Machine learning deals with the problem of extracting features from data so as to solve many different predictive tasks:

• Forecasting (e.g., energy demand prediction, sales)
• Imputing missing data (e.g., Netflix recommendation)
• Detecting anomalies (e.g, intruders, virus mutations)
• Classifying (e.g., credit risk assessment, cancer diagnosis)
• Ranking (e.g., Google search, personalization)
• Summarizing (e.g., Zeit, social media sentiment)
• Decision making (e.g., AI, robotics, trading)

• Human expertise is absent (e.g., navigating on Mars)
• Humans are unable to explain their expertise (e.g., speech recognition, vision, language)
• Solution changes with time (e.g., tracking, temperature control, preferences)
• Solutions need to be adapted to particular case (e.g., biometrics, personalization)
• The problem size is too vast (e.g., calculating web page ranks, matching ads to FB pages)
• Big data - entire Library of Congress info is 20 TB fits into a small portable device; Google process ~24 PB of data/day
• System gets better with more data