-
Notifications
You must be signed in to change notification settings - Fork 3
lostoy/ece285
Folders and files
Name | Name | Last commit message | Last commit date | |
---|---|---|---|---|
Repository files navigation
3.18 Serious bug exists in HW4 Problem3 code!! 3.16 HW5 sol added 3.4: HW4 problem3, added cvx toolbox for lasso. 3.4: HW4 problem3, timing bug and error2 bug fixed 3.1: Added midterm report draft 3.1: Added HW4 problem 3; 2.22: Added HW4 problem1; still working on problem2. 1.26: Serious bugs solved for sol1.m. and added runtime analysis plot. --------HW1---------- --------------flop count analysis------------- A is of size n*m, r is the number of non-zero entries in x. 1) MP: O(m*n*r) 2) OMP: O(m*n*r+n*r*r) 3) LSOMP: O(m*n*r*r) 4) WMP: O(k*n*r) k is the mean of number of cols a_l, such that a_l^T b_{p-1}>t |b_{p-1}|_2 5) Thresholding: O(n*r*r) ---------------Optional Problem--------------- A vast of literature talks about the asymptotic distribution of this problem. We note the original problem is equivalent to finding the pdf of the minimum pair-wise angles of m points uniformly sampled on a n-dimensional unit sphere. We refer to the following journal for a asymptotic distribution of this. Cai, Tony, Jianqing Fan, and Tiefeng Jiang. "Distributions of angles in random packing on spheres." The Journal of Machine Learning Research 14.1 (2013): 1837-1864. Simulation results show that the bias of this asymptotic distribution goes down as m increases. And in extreme of m->\inf, the pdf will be a delta at \mu(A)=1, meaning that \mu(A) will converge to 1 in probability as m goes to infinity. This is intuitively correct, since when we adding more columns into A, the probability of two colinear columns occurs increases. --------HW3----------- See HW3 folder. .\eps_0.001\ is better. (lambda=0.001)
About
No description, website, or topics provided.
Resources
Stars
Watchers
Forks
Releases
No releases published
Packages 0
No packages published