Fast Random Feature Expansions for Nonlinear Regression скачать в хорошем качестве

Fast Random Feature Expansions for Nonlinear Regression

0:01 The theoretical aspects of the random projection method 49:00 The practical part of the tutorial starts. 1:08:53 Explanation of the Fastfood algorithm A tutorial given to the Cambridge Machine Learning Group by David Lopez-Paz http://people.tuebingen.mpg.de/dlopez/ and David Duvenaud http://mlg.eng.cam.ac.uk/duvenaud/ The first half derives the Johnson-Lindenstrauss Lemma, which states that one may randomly project a collection of data points into a lower dimensional space while preserving pairwise point distances. The second half discusses recent developments that have gone even further: non-linear randomised projections can be used to approximate kernel machines and scale them to datasets with millions of features and samples. Slides available at: http://mlg.eng.cam.ac.uk/duvenaud/tal... Q.V. Le, T. Sarlos, A.J. Smola. (2013) Fastfood: Approximating Kernel Expansions in Loglinear Time http://cs.stanford.edu/~quocle/LeSarl...