STAT 535 Midterm: STAT 535 UW Fall13 Solution Midterm13 538
Document Summary
February 19, 2pm february 21, 2013, 11:30am. 17 points (decision regions) (stochastic gradient) (boosting) (convexity) (entropy and information) Exact appr (cid:3)x (cid:3) (cid:3) (cid:3)x (cid:3) (cid:3)x 1-nearest neighbor classi er (cid:3) linear classi er (cid:3) quadratic classi er (cid:3) decision tree (cid:3)x 2 layer neural network a b. Exact appr (cid:3)x (cid:3)x (cid:3)x (cid:3) (cid:3)x (cid:3)x 1-nearest neighbor classi er (cid:3) linear classi er (cid:3) quadratic classi er (cid:3)x decision tree (cid:3) 2 layer neural network. Problem 2 stochastic gradient for additive model. 2. 1 l( ) = 1 i=1 e yi pk=1:m kbk(xi) K e y pj j bj(x) = yj=1:m,j6=k e y j bj e y kbk(x) = ybk(x)l(y, f (x)) and l(y, f (x)) = yb(x)l(y, f (x)) where b(x) is the vector [b1(x) . bm(x)]t . 2. 4 l is convex in because it is a linear combination of the convex functions e t ai with ai = [ yib1(xi) .
