支持向量機(jī)(SVM)實(shí)現(xiàn)的分類算法源碼.doc_第1頁
支持向量機(jī)(SVM)實(shí)現(xiàn)的分類算法源碼.doc_第2頁
支持向量機(jī)(SVM)實(shí)現(xiàn)的分類算法源碼.doc_第3頁
支持向量機(jī)(SVM)實(shí)現(xiàn)的分類算法源碼.doc_第4頁
全文預(yù)覽已結(jié)束

下載本文檔

版權(quán)說明:本文檔由用戶提供并上傳,收益歸屬內(nèi)容提供方,若內(nèi)容存在侵權(quán),請進(jìn)行舉報或認(rèn)領(lǐng)

文檔簡介

支持向量機(jī)(SVM)實(shí)現(xiàn)的分類算法源碼神威 2007-03-12 16:33 發(fā)表function iter, optCond, time, w, gamma = lsvm(A,D,nu,tol,maxIter,alpha, .perturb,normalize);% LSVM Langrangian Support Vector Machine algorithm% LSVM solves a support vector machine problem using an iterative% algorithm inspired by an augmented Lagrangian formulation.% iters = lsvm(A,D)% where A is the data matrix, D is a diagonal matrix of 1s and -1s% indicating which class the points are in, and iters is the number% of iterations the algorithm used.% All the following additional arguments are optional:% iters, optCond, time, w, gamma = .% lsvm(A,D,nu,tol,maxIter,alpha,perturb,normalize)% optCond is the value of the optimality condition at termination.% time is the amount of time the algorithm took to terminate.% w is the vector of coefficients for the separating hyperplane.% gamma is the threshold scalar for the separating hyperplane.% On the right hand side, A and D are required. If the rest are not% specified, the following defaults will be used:% nu = 1/size(A,1), tol = 1e-5, maxIter = 100, alpha = 1.9/nu,% perturb = 0, normalize = 0% perturb indicates that all the data should be perturbed by a random% amount between 0 and the value given. perturb is recommended only% for highly degenerate cases such as the exclusive or.% normalize should be set to 1 if the data should be normalized before% training.% The value -1 can be used for any of the entries (except A and D) to% specify that default values should be used.% Copyright (C) 2000 Olvi L. Mangasarian and David R. Musicant.% Version 1.0 Beta 1% This software is free for academic and research use only.% For commercial use, contact . % If D is a vector, convert it to a diagonal matrix. k,n = size(D); if k=1 | n=1 D=diag(D); end; % Check all components of D and verify that they are +-1 checkall = diag(D)=1 | diag(D)=-1; if any(checkall=0) error(Error in D: classes must be all 1 or -1.); end; m = size(A,1); if exist(nu) | nu=-1 nu = 1/m; end; if exist(tol) | tol=-1 tol = 1e-5; end; if exist(maxIter) | maxIter=-1 maxIter = 100; end; if exist(alpha) | alpha=-1 alpha = 1.9/nu; end; if exist(normalize) | normalize=-1 normalize = 0; end; if exist(perturb) | perturb=-1 perturb = 0; end; % Do a sanity check on alpha if alpha 2/nu, disp(sprintf(Alpha is larger than 2/nu. Algorithm may not converge.); end; % Perturb if appropriate rand(seed,22); if perturb, A = A + rand(size(A)*perturb; end; % Normalize if appropriate if normalize, avg = mean(A); dev = std(A); if (isempty(find(dev=0) A = (A - avg(ones(m,1),:)./dev(ones(m,1),:); else warning(Could not normalize matrix: at least one column is constant.); end; end; % Create matrix H m,n = size(A); e = ones(m,1); H = D*A -e; iter = 0; time = cputime; % K is an intermediate matrix used often in SMW calclulations K = H*inv(speye(n+1)/nu+H*H); % Choose initial value for x x = nu*(1-K*(H*e); % y is the old value for x, used to check for termination y = x + 1; while iter tol % Intermediate calculation which is used twice z = (1+pl(x/nu+H*(H*x)-alpha*x)-1); y = x; % Calculate new value of x x=nu*(z-K*(H*z); iter = iter + 1; end; % Determine outputs time = cputime - time; optCond = norm(x-y); w = A*D*x; gamma = -e*D*x; disp(sprintf(Running time (CPU secs) = %g,time); disp(sprintf(Number of iterations = %d,iter); disp(sprintf(Training accuracy = %

溫馨提示

  • 1. 本站所有資源如無特殊說明,都需要本地電腦安裝OFFICE2007和PDF閱讀器。圖紙軟件為CAD,CAXA,PROE,UG,SolidWorks等.壓縮文件請下載最新的WinRAR軟件解壓。
  • 2. 本站的文檔不包含任何第三方提供的附件圖紙等,如果需要附件,請聯(lián)系上傳者。文件的所有權(quán)益歸上傳用戶所有。
  • 3. 本站RAR壓縮包中若帶圖紙,網(wǎng)頁內(nèi)容里面會有圖紙預(yù)覽,若沒有圖紙預(yù)覽就沒有圖紙。
  • 4. 未經(jīng)權(quán)益所有人同意不得將文件中的內(nèi)容挪作商業(yè)或盈利用途。
  • 5. 人人文庫網(wǎng)僅提供信息存儲空間,僅對用戶上傳內(nèi)容的表現(xiàn)方式做保護(hù)處理,對用戶上傳分享的文檔內(nèi)容本身不做任何修改或編輯,并不能對任何下載內(nèi)容負(fù)責(zé)。
  • 6. 下載文件中如有侵權(quán)或不適當(dāng)內(nèi)容,請與我們聯(lián)系,我們立即糾正。
  • 7. 本站不保證下載資源的準(zhǔn)確性、安全性和完整性, 同時也不承擔(dān)用戶因使用這些下載資源對自己和他人造成任何形式的傷害或損失。

評論

0/150

提交評論