神经网络识别手写数字（含数据和代码，只要有matlab就可以运行）

function [X, fX, i] = fmincg(f, X, options, hidden_layer_size, input_layer_size, class_num, X_test, Y_test, P1, P2, P3, P4, P5) if exist('options', 'var') && ~isempty(options) && isfield(options, 'MaxIter') length = options.MaxIter; else length = 100; end rho = 0.01; sig = 0.5; INT = 0.1; EXT = 3.0; MAX = 20; RATIO = 100; argstr = ['feval(f, X']; for i = 1:(nargin - 8) argstr = [argstr, ',P', int2str(i)]; end argstr = [argstr, ')']; if max(size(length)) == 2, red=length(2); length=length(1); else red=1; end i = 0; failed = 0; fX = []; [f1 df1] = eval(argstr); i = i + (length<0); s = -df1; d1 = -s'*s; z1 = red/(1-d1); while i < abs(length) i = i + (length>0); X0 = X; f0 = f1; df0 = df1; X = X + z1*s; [f2 df2] = eval(argstr); i = i + (length<0); d2 = df2'*s; f3 = f1; d3 = d1; z3 = -z1; if length>0, M = MAX; else M = min(MAX, -length-i); end success = 0; limit = -1; while 1 while ((f2 > f1+z1*rho*d1) || (d2 > -sig*d1)) && (M > 0) limit = z1; if f2 > f1 z2 = z3 - (0.5*d3*z3*z3)/(d3*z3+f2-f3); else A = 6*(f2-f3)/z3+3*(d2+d3); B = 3*(f3-f2)-z3*(d3+2*d2); z2 = (sqrt(B*B-A*d2*z3*z3)-B)/A; end if isnan(z2) || isinf(z2) z2 = z3/2; end z2 = max(min(z2, INT*z3),(1-INT)*z3); z1 = z1 + z2; X = X + z2*s; [f2 df2] = eval(argstr); M = M - 1; i = i + (length<0); d2 = df2'*s; z3 = z3-z2; end if f2 > f1+z1*rho*d1 || d2 > -sig*d1 break; elseif d2 > sig*d1 success = 1; break; elseif M == 0 break; end A = 6*(f2-f3)/z3+3*(d2+d3); B = 3*(f3-f2)-z3*(d3+2*d2); z2 = -d2*z3*z3/(B+sqrt(B*B-A*d2*z3*z3)); if ~isreal(z2) || isnan(z2) || isinf(z2) || z2 < 0 if limit < -0.5 z2 = z1 * (EXT-1); else z2 = (limit-z1)/2; end elseif (limit > -0.5) && (z2+z1 > limit) z2 = (limit-z1)/2; elseif (limit < -0.5) && (z2+z1 > z1*EXT) z2 = z1*(EXT-1.0); z2 = -z3*INT; elseif (limit > -0.5) && (z2 < (limit-z1)*(1.0-INT)) z2 = (limit-z1)*(1.0-INT); end f3 = f2; d3 = d2; z3 = -z2; z1 = z1 + z2; X = X + z2*s; [f2 df2] = eval(argstr); M = M - 1; i = i + (length<0); d2 = df2'*s; end if success f1 = f2; fX = [fX' f1]'; nn_params = X; Theta1 = reshape(nn_params(1:hidden_layer_size * (input_layer_size + 1)), ... hidden_layer_size, (input_layer_size + 1)); Theta2 = reshape(nn_params((1 + (hidden_layer_size * (input_layer_size + 1))):end), ... class_num, (hidden_layer_size + 1)); plotnumData(Theta1(:, 2:end)); hold on title(strcat('迭代步数: ',num2str(i))) pred = predict(Theta1, Theta2, X_test); fprintf('迭代步数 %4i |***| Cost值 %4.6e |***| 测试集准确率: %2.2f\r', i, f1,mean(double(pred == Y_test)) * 100); s = (df2'*df2-df1'*df2)/(df1'*df1)*s - df2; tmp = df1; df1 = df2; df2 = tmp; d2 = df1'*s; if d2 > 0 s = -df1; d2 = -s'*s; end z1 = z1 * min(RATIO, d1/(d2-realmin)); d1 = d2; failed = 0; else X = X0; f1 = f0; df1 = df0; if failed || i > abs(length) break; end tmp = df1; df1 = df2; df2 = tmp; s = -df1; d1 = -s'*s; z1 = 1/(1-d1); failed = 1; end end end