This preview shows page 1-2-3-4-5-6-7-8-54-55-56-57-58-59-60-109-110-111-112-113-114-115-116 out of 116 pages.
Statistical Learning Theory and ApplicationsClass Times:Monday and Wednesday 10:30-12:00Units: 3-0-9 H,GLocation:46-5193Instructors:Tomaso Poggio (TP), Lorenzo Rosasco (LR), Charlie Frogner (CF), Guille D. Canas (GJ)Office Hours:Friday 1-2 pm in 46-5156, CBCL loungeEmail Contact :[email protected] in 2012 Saturday, February 4, 2012Rules of the game: • problem sets (3, last one consists of posting Wikipedia article)• final project (between review and j. paper): you have to give us title+abstract before March 29th• scribing• participation• Grading is based on Psets (20%+20%+10%) + Final Project (30%) + Scribing (10%) + Participation (10%)Slides on the Web siteStaff mailing list is [email protected] Student list will be [email protected] Please fill form!send email to us if you want to be added to mailing listClasshttp://www.mit.edu/~9.520/Saturday, February 4, 2012Mathcamps (optional): • Functional analysis (~45mins)• Probability (~45mins)Feb 13 7pm-9pm???Classhttp://www.mit.edu/~9.520/Saturday, February 4, 2012We will provide some in the next few classes and we will speak more about them just before spring breakStatistical Learning Theory and Applications:ProjectsSaturday, February 4, 20125Projects 2011Final writeup due Sunday, May 15th, by midnight• Project Ideas Contact: Instructors • Part-based Human Recognition in Videos Contact: Hueihan Jhuang • Solving Large Scale Kernel Machines using Random Features Contact: Nicholas Edelman • Evaluating which Classifiers Work Best for Decoding Neural Data Contact: Ethan Meyers • Does learning from segmented images aid categorization? Contact: Cheston Tan• What can humans see with a single glance? Contact: Cheston Tan• Demo of the motion silencing effect Contact: Cheston Tan• When invariance learning goes wrong Contact: Joel Leibo• More TBAWe will provide some in the next few classes and we will speak more about them before spring breakStatistical Learning Theory and Applications:ProjectsSaturday, February 4, 20129.520 Statistical Learning Theory and Applications Class 26: Project presentations (past examples)2:35-2:50 "Learning card playing strategies with SVMs", David Craft and Timothy Chan2:50-3:00 "Artificial Markets: Learning to trade using Support Vector Machines“, Adlar Kim3:00-3:10 "Feature selection: literature review and new development'‘, Wei Wu3:10—3:25 "Man vs machines: A computational study on face detection" Thomas SerreSaturday, February 4, 20129.520 Statistical Learning Theory and Applications (2007) 10:30 - Simon Laflamme “Online Learning Algorithm for Structural Control using Magnetorheological Actuators” - Emily Shen “Time series prediction” - Zak Stone “Facebook project” - Jeff Miller “Clustering features in the standard model of cortex” - Manuel Rivas "Learning Age from Gene Expression Data“ - Demba Ba “Sparse Approximation of the Spectrogram via Matching Pursuits: Applications to Speech Analysis”- Nikon Rasumov "Data mining in controlled environment and real data"Saturday, February 4, 20128• Reviews of a topic.• Projects, simulations and/or theorems:-- Learning to rank papers/grants: replacing review panels-- Oscillations and iterations in optimization-- Class-specific computations and architecture of recognition-- Sparseness and recall from visual associative memory-- The surprising usefulness of sloppy arithmetic: study of bits and their tradeoff in hierarchical architecturesClass projects: examplesSaturday, February 4, 20129Class projects: an exampleProject: prove theorem!Saturday, February 4, 2012Problem Set 3: posting/editing article on Wikipedia10• Computational learning theory: to be redone or new entry in Generalization Bounds• RKHS is ok but could be improved on the learning side• Stability in Learning Theory (batch and online) is missing • Radial basis function network should be rewritten or edited • VC theory exists in a minimalistic form• Regularization networks/theory IS TERRIBLE...EASY TO IMPROVE• Statistical learning theory is a messSaturday, February 4, 2012Overview of overview•Context for this course: a golden age for new AI and the key role of Machine Learning•Success stories from past research in Machine Learning: examples of engineering applications •Statistical Learning Theory•A new cycle of basic research on learning: computer science and neuroscience, learning and the brain•A Center for Brains, Minds and MachinesSaturday, February 4, 201212Saturday, February 4, 2012Overview of overview•Context for this course: a golden age for new AI and the key role of Machine Learning•Success stories from past research in Machine Learning: examples of engineering applications •Statistical Learning Theory•A new cycle of basic research on learning: computer science and neuroscience, learning and the brain•A Center for Brains, Minds and MachinesSaturday, February 4, 2012The$problem$of$intelligence$is$one$of$the$great$problems$in$science,$probably$the$greatest.Research$on$intelligence:$•$a$great$intellectual$mission•$will$help$cure$mental$diseases$and$develop$more$intelligent$ar<facts$•$will$improve$the$mechanisms$for$collec<ve$decisionsThese$advances$will$be$cri<cal$to$of$ our$soci ety’s•$future$prosperity•$educa<on,$$health,$$security$The problem of intelligence: how it arises in the brain and how to replicate it in machinesSaturday, February 4, 2012National priorities: grand challenges for 21st century engineering (NAE)• Fully half (7 out of 14) focus on the frontiers of intelligence:– Reverse engineer the brain– Advance personalized learning– Enhance virtual reality– Engineer the tools of scientific discovery– Advance health informatics– Engineer better medicines– Secure cyberspaceSaturday, February 4, 2012At the core of the problem of Intelligence is the problem of LearningLearning is the gateway to understanding the brain and to making intelligent machines. Problem of learning: a focus for o math o computer algorithms o neuroscience Saturday, February 4, 2012LEARNING THEORY+ ALGORITHMSCOMPUTATIONAL NEUROSCIENCE:models+experimentsENGINEERING APPLICATIONSPoggio, T. and F. Girosi. Networks for Approximation and Learning, Proceedings of the IEEE 1990) also Science, 1990Poggio, T. and S.Smale. The Mathematics ofLearning: DealingwithData, Notices American Mathematical Society (AMS), 2003Poggio, T., R. Rifkin, S. Mukherjee and P.
View Full Document