- ignore *.asv

git-svn-id: http://moon:8086/svn/matlab/trunk@96 801c6759-fa7c-4059-a304-17956f83a07c
2016-07-12 20:36:03 +00:00
parent 32063406bd
commit 79b4784a01
8 changed files with 0 additions and 208 deletions
@@ -1,20 +0,0 @@
-function [xpca xzca] = PCAwhite(x)
-size(x)
-x = x';
-k= 10;
-epsilon=1e-42;
-avg = mean(x, 1)
-
-x = x - repmat(avg, size(x, 1), 1);
-
-sigma = x * x' / size(x, 2);
-size(sigma)
-[U,S,V] = svd(sigma);
-%xRot = U' * x;          % rotated version of the data. 
-%xTilde = U(:,1:k)' * x; % reduced dimension representation of the data, 
-                         % where k is the number of eigenvectors to keep
-
-xpca = diag(1./sqrt(diag(S) + epsilon)) * U' * x;
-xzca = U * diag(1./sqrt(diag(S) + epsilon)) * U' * x;
-
-plot(x(
@@ -1,12 +0,0 @@
-function jbatch2()
-numLabels = 10;
-
-maxNumSamples = -1;
-di
-for d=1:numLabels,
-  Dall(d) = load(['digit' num2str(d-1)]);
-  fprintf('%5d Digits of class %d\n',size(Dall(d).D,1),d-1);
-  maxNumSamples = max(maxNumSamples, size(Dall(d).D,1));
-  jconv(Dall(d).D/256, d-1);
-%  jimage(Dall{d}/256, d-1, 28, 28);
-end;
@@ -1,13 +0,0 @@
-function jconv(data, label)
-[numTrain, numPixel] = size(data);
-
-s = ['mnist_' num2str(label) '.trainingStates.dat']
-fid = fopen(s, 'w');
-fprintf(fid, '%d\n', numTrain);
-
-for m=1:numTrain,
-	jwrite(fid, data(m,:));
-end
-
-
-fclose(fid);
@@ -1,12 +0,0 @@
-function jimage(data, label, nx, ny)
-[numTrain, numPixel] = size(data);
-
-for m=1:numTrain,
-    im0 = reshape(data(m,:), nx, ny)';
-    imwrite(im0, ['mnist_' num2str(label) '_' num2str(m) '.tif']);
-    [im_on, im_off] = retina(im0, 7, 0.3);
-end
-
-%imwrite(im_on, [dst_dir 'images\' filename '.on.tif']);
-%imwrite(im_off, [dst_dir 'images\' filename '.off.tif']);
-%imwrite(0.5*(im_on - im_off) + 0.5, [dst_dir 'images\' filename '.on_off.tif']);
@@ -1,13 +0,0 @@
-function jwrite_retina(fid, data, nx, ny)
-
-numPixel = length(data);
-
-fprintf(fid, '%d\n', numPixel);
-im0 = reshape(data(m,:), nx, ny)';
-    imwrite(im0, ['mnist_' num2str(label) '_' num2str(m) '.tif']);
-    [im_on, im_off] = retina(im0, 7, 0.3);
-    imwrite(0.5*(im_on - im_off) + 0.5,  ['mnist_retina_' num2str(label) '_' num2str(m) '.tif']);
-
-for n=1:numPixel,
-	fprintf(fid, '%f\n', data(n));
-end
@@ -1,19 +0,0 @@
-function norbload()
-
-fid=fopen('smallnorb-5x46789x9x18x6x2x96x96-training-dat.mat','r');
-fread(fid,4,'uchar'); % result = [85 76 61 30], byte matrix(in base 16: [55 4C 3D 1E])
-fread(fid,4,'uchar'); % result = [4 0 0 0], ndim = 4
-numImages = fread(fid,4,'uchar'); % result = [236 94 0 0], dim0 = 24300 (=94*256+236)
-fread(fid,4,'uchar'); % result = [2 0 0 0], dim1 = 2
-fread(fid,4,'uchar'); % result = [96 0 0 0], dim2 = 96
-fread(fid,4,'uchar'); % result = [96 0 0 0], dim3 = 96
-
-numImages = numImages
-for n=1:200,
-    im = transpose(reshape(fread(fid,96*96),96,96))/256;
-    if (mod(n, 2) == 0)
-        size(im)
-        imshow(im); % show the first image
-        pause (0.1);
-    end
-end;
@@ -1,43 +0,0 @@
-function pca_eval()
-
-close all;
-
-Ntrain = 100;
-Ndim = 2;
-K = 1;
-epsilon = 0.1;
-
-x = 2*(rand(Ndim, Ntrain)-0.5);
-x = 0.3*x + repmat(linspace(-0.7,0.7, Ntrain), Ndim, 1);
-plot(x(1,:), x(2,:), '+'); axis([-1 1 -1 1]); grid; title('Raw data'); xlabel('x_1'); ylabel('x_2');
-
-sigma = x * x' / size(x, 2)
-[U,S,V] = svd(sigma)
-xrot = U' * x;
-figure;
-plot(xrot(1,:), xrot(2,:), '+'); axis([-1 1 -1 1]); grid; title('Rotated data'); xlabel('x_{rot,1}'); ylabel('x_{rot,2}');
-
-xpca = pcdiag(sqrt(1./(diag(S) + epsilon))) * xrot;
-figure;
-plot(xpca(1,:), xpca(2,:), '+'); axis([-1 1 -1 1]); grid; title('Whitened data dim=2'); xlabel('x_{pca,1}'); ylabel('x_{pca,2}');
-
-
-xtilde = zeros(size(xrot));
-xtilde(1:K, :) = xrot(1:K, :);
-
-lambda = diag(S);
-retained_variance = sum(lambda(1:K))./sum(lambda)
-
-figure;
-plot(xtilde(1,:), xtilde(2,:), '+'); axis([-1 1 -1 1]); grid; title('Rotated data reduced to dim = 1'); xlabel('x_{tilde,1}'); ylabel('x_{tilde,2}');
-
-size (U)
-size(xrot)
-xhat = U(:, 1:K) * xrot(1:K, :);
-
-figure;
-plot(xhat(1,:), xhat(2,:), '+'); axis([-1 1 -1 1]); grid; title('Data reduced to dim = 1'); xlabel('x_{hat,1}'); ylabel('x_{hat,2}');
-
-xpca1 = diag(sqrt(1./(diag(S)))) * xhat;
-figure;
-plot(xpca1(1,:), xpca1(2,:), '+'); axis([-1 1 -1 1]); grid; title('Whitened data dim=1'); xlabel('x_{pca,1}'); ylabel('x_{pca,2}');
@@ -1,76 +0,0 @@
-function rbm_eval(nh, batch, numEpochs)
-
-[nv batchsize] = size(batch);
-
-numGibbs = 1;
-muW = 0.1;
-muBV = 0.1;
-muBH = 0.1;
-momentum = 0.5;
-momentum_final = 0.9;
-weightDecay = 0.000;
-
-w = 0.1*randn(nv, nh);
-bv = zeros(nv, 1);
-bh = zeros(1, nh);
-
-v = zeros(batchsize, nv);
-h = zeros(batchsize, nh);
-
-diffW = zeros(nv, nh);
-diffBV = zeros(1, nv);
-diffBH = zeros(1, nh);
-diffErr = zeros(nv, batchsize);
-
-dW = zeros(nv, nh);
-dBV = zeros(nv, 1);
-dBH = zeros(nh, 1);
-
-for k=1:numEpochs
-    errsum = 0;
-    % Positive phase
-    v = batch;
-    h = 1./(1 + exp(-(v' * w + repmat(bh, batchsize, 1))));
-    
-    % Update difference
-    diffW = v * h;
-    diffBV = sum(v, 2);
-    diffBH = sum(h')';
-    diffErr = v;
-    % Negative phase
-    for n=1:numGibbs
-        % h is sampled
-        h = h > rand(batchsize, nh);
-
-        % reconstruct
-        v = 1./(1 + exp(-(h *w'+ repmat(bv, batchsize, 1))));
-        
-        % Get new h probabilities from reconstruction
-        h = 1./(1 + exp(-(v' * w + repmat(bh, batchsize, 1))))';
-    end
-    % Update difference
-    diffW = diffW - v * h';
-    diffBV = diffBV - sum(v, 2);
-    diffBH = diffBH - sum(h');
-    diffErr = diffErr - v;
-
-    err = sum(sum((diffErr).^2));
-    errsum = errsum + err;
-    
-    if k > 5,
-        momentum = momentum_final;
-    end;
-    
-    % Update parameter gradients
-    dW = momentum*dW + muW*(diffW/batchsize - weightDecay*w);
-    dBV = momentum*dBV + muBV*diffBV/batchsize;
-    dBH = momentum*dBH + muBH*diffBH/batchsize;
-
-    % Update parameters
-    w = w + dW;
-    bv = bv + dBV;
-    bh = bh + dBH;
-end
-v = v
-err = err
-errsum = errsum