tailieunhanh - Challenges in Machine Learning and Data Mining

Challenges in Machine Learning and Data Mining presents about generative vs. discriminative learning, learning from non-vectorial data, Beyond classification and regression, Distributed data mining, Machine learning bottlenecks, Intelligible models, Combining learning methods. | Challenges in Machine Learning and Data Mining Tu Bao Ho JAIST Based on materials from 1. 9 challenges in ML Caruana Joachims 2. 10 challenging problems in DM Yang Wu in International Journal of Information Technology Decision Making 2006 What is machine learning The goal of machine learning is to build computer systems that can adapt and learn from their experience Tom Dietterich . A computer program is said to learn from experience E with respect to some class of tasks T and performance measure p if its performance at tasks in T as measure by p improves with experience Tom Mitchell book p. 2 . ML problems can be formulated as Given xI yP x2 y2 . xn yn - Xị is description of an object phenomenon etc. - Yi is some property of Xị if not available learning is unsupervised Find a function f x that f Xj - Yj Finding hypothesis f in a huge hypothesis space F by narrowing the search with constraints bias Machine learning and data mining Machine learning To build computer systems that learn as well as human does science of learning from data . ICML since 1982 23th ICMLin 2006 ECML since 1989. ECML PKDD since 2001. ACML starts Nov. 2009. Data minin To find new and useful knowledge from large datasets data engineering . ACMSIGKDD since 1995 PKDD and PAKDD since 1997 IEEE ICDM and SIAM DM since 2000 etc. Note Difference between statistics machine learning data mining Co-chair of steering Committee of PAKDD member of steering Committee of ACML 2 Overview of ML challenges 1. Generative vs. discriminative learning 2. Learning from non-vectorial data 3. Beyond classification and regression 4. Distributed data mining 5. Machine learning bottlenecks 6. Intelligible models 7. Combining learning methods 8. Unsupervised learning comes of age 9. More informed information access 1. Generative vs. discriminative methods Training classifiers involves estimating f X Y or P Y X . Examples P apple I red A round P noun I cá Generative classifiers Assume some functional form for p x Y P Y .