基于SVM支持向量机的数据二分类matlab仿真,SVM自己编程实现不使用matlab工具箱+代码仿真操作视频

function alpha = svm(K, C, params) alpha = zeros(size(K, 1), 1); % Set alpha0 to be C / 2, which will be optimum if mu is very large alpha = C / 2; % Set mu0 to be the average of the diagonal of K mu = 1000*mean( diag( K ) ); % Set the beta parameter beta = 0.667; % Get the parameters epsilon = params.epsilon; Nmax = params.maxiter; b = params.barrier; % set up one matrix one = ones(size(alpha)); for i = 1:Nmax % Find the gradient g = 0.5*(K + K')*alpha - one - mu*alpha.^(-1) + mu*(C - alpha).^(-1); % Find the Hessian H = 0.5*(K + K') + diag(mu*(alpha.^(-1)).*(alpha.^(-1)) + mu*((C - alpha).^(-1)).*((C - alpha).^(-1))); % Compute the Newton direction d = -(H \ g); % update iterators %i = i + 1; t = 1; % Perform backtracking line search while (f(alpha + t*d, C, K, mu) > f(alpha, C, K, mu) + 0.5*t*g.'*d) t = t * beta; end if( t > 0.99 && mu > epsilon ) mu = mu * b; end alpha = alpha + t * d; if( norm(d) <= epsilon && mu <= epsilon ) break; end end end