Commit 3.2

Contents.m

@@ -0,0 +1,174 @@
+% Netlab Toolbox
+% Version 3.2.1 	 31-Oct-2001
+%
+% conffig  -  Display a confusion matrix. 
+% confmat  -  Compute a confusion matrix. 
+% conjgrad -  Conjugate gradients optimization. 
+% consist  -  Check that arguments are consistent. 
+% datread  -  Read data from an ascii file. 
+% datwrite -  Write data to ascii file. 
+% dem2ddat -  Generates two dimensional data for demos. 
+% demard   -  Automatic relevance determination using the MLP. 
+% demev1   -  Demonstrate Bayesian regression for the MLP. 
+% demev2   -  Demonstrate Bayesian classification for the MLP. 
+% demev3   -  Demonstrate Bayesian regression for the RBF. 
+% demgauss -  Demonstrate sampling from Gaussian distributions. 
+% demglm1  -  Demonstrate simple classification using a generalized linear model. 
+% demglm2  -  Demonstrate simple classification using a generalized linear model. 
+% demgmm1  -  Demonstrate density modelling with a Gaussian mixture model. 
+% demgmm3  -  Demonstrate density modelling with a Gaussian mixture model. 
+% demgmm4  -  Demonstrate density modelling with a Gaussian mixture model. 
+% demgmm5  -  Demonstrate density modelling with a PPCA mixture model. 
+% demgp    -  Demonstrate simple regression using a Gaussian Process. 
+% demgpard -  Demonstrate ARD using a Gaussian Process. 
+% demgpot  -  Computes the gradient of the negative log likelihood for a mixture model. 
+% demgtm1  -  Demonstrate EM for GTM. 
+% demgtm2  -  Demonstrate GTM for visualisation. 
+% demhint  -  Demonstration of Hinton diagram for 2-layer feed-forward network. 
+% demhmc1  -  Demonstrate Hybrid Monte Carlo sampling on mixture of two Gaussians. 
+% demhmc2  -  Demonstrate Bayesian regression with Hybrid Monte Carlo sampling. 
+% demhmc3  -  Demonstrate Bayesian regression with Hybrid Monte Carlo sampling. 
+% demkmean -  Demonstrate simple clustering model trained with K-means. 
+% demknn1  -  Demonstrate nearest neighbour classifier. 
+% demmdn1  -  Demonstrate fitting a multi-valued function using a Mixture Density Network. 
+% demmet1  -  Demonstrate Markov Chain Monte Carlo sampling on a Gaussian. 
+% demmlp1  -  Demonstrate simple regression using a multi-layer perceptron 
+% demmlp2  -  Demonstrate simple classification using a multi-layer perceptron 
+% demnlab  -  A front-end Graphical User Interface to the demos 
+% demns1   -  Demonstrate Neuroscale for visualisation. 
+% demolgd1 -  Demonstrate simple MLP optimisation with on-line gradient descent 
+% demopt1  -  Demonstrate different optimisers on Rosenbrock's function. 
+% dempot   -  Computes the negative log likelihood for a mixture model. 
+% demprgp  -  Demonstrate sampling from a Gaussian Process prior. 
+% demprior -  Demonstrate sampling from a multi-parameter Gaussian prior. 
+% demrbf1  -  Demonstrate simple regression using a radial basis function network. 
+% demsom1  -  Demonstrate SOM for visualisation. 
+% demtrain -  Demonstrate training of MLP network. 
+% dist2    -  Calculates squared distance between two sets of points. 
+% eigdec   -  Sorted eigendecomposition 
+% errbayes -  Evaluate Bayesian error function for network. 
+% evidence -  Re-estimate hyperparameters using evidence approximation. 
+% fevbayes -  Evaluate Bayesian regularisation for network forward propagation. 
+% gauss    -  Evaluate a Gaussian distribution. 
+% gbayes   -  Evaluate gradient of Bayesian error function for network. 
+% glm      -  Create a generalized linear model. 
+% glmderiv -  Evaluate derivatives of GLM outputs with respect to weights. 
+% glmerr   -  Evaluate error function for generalized linear model. 
+% glmevfwd -  Forward propagation with evidence for GLM 
+% glmfwd   -  Forward propagation through generalized linear model. 
+% glmgrad  -  Evaluate gradient of error function for generalized linear model. 
+% glmhess  -  Evaluate the Hessian matrix for a generalised linear model. 
+% glminit  -  Initialise the weights in a generalized linear model. 
+% glmpak   -  Combines weights and biases into one weights vector. 
+% glmtrain -  Specialised training of generalized linear model 
+% glmunpak -  Separates weights vector into weight and bias matrices. 
+% gmm      -  Creates a Gaussian mixture model with specified architecture. 
+% gmmactiv -  Computes the activations of a Gaussian mixture model. 
+% gmmem    -  EM algorithm for Gaussian mixture model. 
+% gmminit  -  Initialises Gaussian mixture model from data 
+% gmmpak   -  Combines all the parameters in a Gaussian mixture model into one vector. 
+% gmmpost  -  Computes the class posterior probabilities of a Gaussian mixture model. 
+% gmmprob  -  Computes the data probability for a Gaussian mixture model. 
+% gmmsamp  -  Sample from a Gaussian mixture distribution. 
+% gmmunpak -  Separates a vector of Gaussian mixture model parameters into its components. 
+% gp       -  Create a Gaussian Process. 
+% gpcovar  -  Calculate the covariance for a Gaussian Process. 
+% gpcovarf -  Calculate the covariance function for a Gaussian Process. 
+% gpcovarp -  Calculate the prior covariance for a Gaussian Process. 
+% gperr    -  Evaluate error function for Gaussian Process. 
+% gpfwd    -  Forward propagation through Gaussian Process. 
+% gpgrad   -  Evaluate error gradient for Gaussian Process. 
+% gpinit   -  Initialise Gaussian Process model. 
+% gppak    -  Combines GP hyperparameters into one vector. 
+% gpunpak  -  Separates hyperparameter vector into components. 
+% gradchek -  Checks a user-defined gradient function using finite differences. 
+% graddesc -  Gradient descent optimization. 
+% gsamp    -  Sample from a Gaussian distribution. 
+% gtm      -  Create a Generative Topographic Map. 
+% gtmem    -  EM algorithm for Generative Topographic Mapping. 
+% gtmfwd   -  Forward propagation through GTM. 
+% gtminit  -  Initialise the weights and latent sample in a GTM. 
+% gtmlmean -  Mean responsibility for data in a GTM. 
+% gtmlmode -  Mode responsibility for data in a GTM. 
+% gtmmag   -  Magnification factors for a GTM 
+% gtmpost  -  Latent space responsibility for data in a GTM. 
+% gtmprob  -  Probability for data under a GTM. 
+% hbayes   -  Evaluate Hessian of Bayesian error function for network. 
+% hesschek -  Use central differences to confirm correct evaluation of Hessian matrix. 
+% hintmat  -  Evaluates the coordinates of the patches for a Hinton diagram. 
+% hinton   -  Plot Hinton diagram for a weight matrix. 
+% histp    -  Histogram estimate of 1-dimensional probability distribution. 
+% hmc      -  Hybrid Monte Carlo sampling. 
+% kmeans   -  Trains a k means cluster model. 
+% knn      -  Creates a K-nearest-neighbour classifier. 
+% knnfwd   -  Forward propagation through a K-nearest-neighbour classifier. 
+% linef    -  Calculate function value along a line. 
+% linemin  -  One dimensional minimization. 
+% mdn      -  Creates a Mixture Density Network with specified architecture. 
+% mdn2gmm  -  Converts an MDN mixture data structure to array of GMMs. 
+% mdndist2 -  Calculates squared distance between centres of Gaussian kernels and data 
+% mdnerr   -  Evaluate error function for Mixture Density Network. 
+% mdnfwd   -  Forward propagation through Mixture Density Network. 
+% mdngrad  -  Evaluate gradient of error function for Mixture Density Network. 
+% mdninit  -  Initialise the weights in a Mixture Density Network. 
+% mdnpak   -  Combines weights and biases into one weights vector. 
+% mdnpost  -  Computes the posterior probability for each MDN mixture component. 
+% mdnprob  -  Computes the data probability likelihood for an MDN mixture structure. 
+% mdnunpak -  Separates weights vector into weight and bias matrices. 
+% metrop   -  Markov Chain Monte Carlo sampling with Metropolis algorithm. 
+% minbrack -  Bracket a minimum of a function of one variable. 
+% mlp      -  Create a 2-layer feedforward network. 
+% mlpbkp   -  Backpropagate gradient of error function for 2-layer network. 
+% mlpderiv -  Evaluate derivatives of network outputs with respect to weights. 
+% mlperr   -  Evaluate error function for 2-layer network. 
+% mlpevfwd -  Forward propagation with evidence for MLP 
+% mlpfwd   -  Forward propagation through 2-layer network. 
+% mlpgrad  -  Evaluate gradient of error function for 2-layer network. 
+% mlphdotv -  Evaluate the product of the data Hessian with a vector. 
+% mlphess  -  Evaluate the Hessian matrix for a multi-layer perceptron network. 
+% mlphint  -  Plot Hinton diagram for 2-layer feed-forward network. 
+% mlpinit  -  Initialise the weights in a 2-layer feedforward network. 
+% mlppak   -  Combines weights and biases into one weights vector. 
+% mlpprior -  Create Gaussian prior for mlp. 
+% mlptrain -  Utility to train an MLP network for demtrain 
+% mlpunpak -  Separates weights vector into weight and bias matrices. 
+% netderiv -  Evaluate derivatives of network outputs by weights generically. 
+% neterr   -  Evaluate network error function for generic optimizers 
+% netevfwd -  Generic forward propagation with evidence for network 
+% netgrad  -  Evaluate network error gradient for generic optimizers 
+% nethess  -  Evaluate network Hessian 
+% netinit  -  Initialise the weights in a network. 
+% netopt   -  Optimize the weights in a network model. 
+% netpak   -  Combines weights and biases into one weights vector. 
+% netunpak -  Separates weights vector into weight and bias matrices. 
+% olgd     -  On-line gradient descent optimization. 
+% pca      -  Principal Components Analysis 
+% plotmat  -  Display a matrix. 
+% ppca     -  Probabilistic Principal Components Analysis 
+% quasinew -  Quasi-Newton optimization. 
+% rbf      -  Creates an RBF network with specified architecture 
+% rbfbkp   -  Backpropagate gradient of error function for RBF network. 
+% rbfderiv -  Evaluate derivatives of RBF network outputs with respect to weights. 
+% rbferr   -  Evaluate error function for RBF network. 
+% rbfevfwd -  Forward propagation with evidence for RBF 
+% rbffwd   -  Forward propagation through RBF network with linear outputs. 
+% rbfgrad  -  Evaluate gradient of error function for RBF network. 
+% rbfhess  -  Evaluate the Hessian matrix for RBF network. 
+% rbfjacob -  Evaluate derivatives of RBF network outputs with respect to inputs. 
+% rbfpak   -  Combines all the parameters in an RBF network into one weights vector. 
+% rbfprior -  Create Gaussian prior and output layer mask for RBF. 
+% rbfsetbf -  Set basis functions of RBF from data. 
+% rbfsetfw -  Set basis function widths of RBF. 
+% rbftrain -  Two stage training of RBF network. 
+% rbfunpak -  Separates a vector of RBF weights into its components. 
+% rosegrad -  Calculate gradient of Rosenbrock's function. 
+% rosen    -  Calculate Rosenbrock's function. 
+% scg      -  Scaled conjugate gradient optimization. 
+% som      -  Creates a Self-Organising Map. 
+% somfwd   -  Forward propagation through a Self-Organising Map. 
+% sompak   -  Combines node weights into one weights matrix. 
+% somtrain -  Kohonen training algorithm for SOM. 
+% somunpak -  Replaces node weights in SOM. 
+%
+%	Copyright (c) Ian T Nabney (1996-2001)
+%

conffig.m

@@ -0,0 +1,29 @@
+function fh=conffig(y, t)
+%CONFFIG Display a confusion matrix.
+%
+%	Description
+%	CONFFIG(Y, T) displays the confusion matrix  and classification
+%	performance for the predictions mat{y} compared with the targets T.
+%	The data is assumed to be in a 1-of-N encoding, unless there is just
+%	one column, when it is assumed to be a 2 class problem with a 0-1
+%	encoding.  Each row of Y and T corresponds to a single example.
+%
+%	In the confusion matrix, the rows represent the true classes and the
+%	columns the predicted classes.
+%
+%	FH = CONFFIG(Y, T) also returns the figure handle FH which  can be
+%	used, for instance, to delete the figure when it is no longer needed.
+%
+%	See also
+%	CONFMAT, DEMTRAIN
+%
+
+%	Copyright (c) Ian T Nabney (1996-2001)
+
+[C, rate] = confmat(y, t);
+
+fh = figure('Name', 'Confusion matrix', ...
+  'NumberTitle', 'off');
+
+plotmat(C, 'k', 'k', 14);
+title(['Classification rate: ' num2str(rate(1)) '%'], 'FontSize', 14);

confmat.m

@@ -0,0 +1,56 @@
+function [C,rate]=confmat(Y,T)
+%CONFMAT Compute a confusion matrix.
+%
+%	Description
+%	[C, RATE] = CONFMAT(Y, T) computes the confusion matrix C and
+%	classification performance RATE for the predictions mat{y} compared
+%	with the targets T.  The data is assumed to be in a 1-of-N encoding,
+%	unless there is just one column, when it is assumed to be a 2 class
+%	problem with a 0-1 encoding.  Each row of Y and T corresponds to a
+%	single example.
+%
+%	In the confusion matrix, the rows represent the true classes and the
+%	columns the predicted classes.  The vector RATE has two entries: the
+%	percentage of correct classifications and the total number of correct
+%	classifications.
+%
+%	See also
+%	CONFFIG, DEMTRAIN
+%
+
+%	Copyright (c) Ian T Nabney (1996-2001)
+
+[n c]=size(Y);
+[n2 c2]=size(T);
+
+if n~=n2 | c~=c2
+  error('Outputs and targets are different sizes')
+end
+
+if c > 1
+  % Find the winning class assuming 1-of-N encoding
+  [maximum Yclass] = max(Y', [], 1);
+
+  TL=[1:c]*T';
+else
+  % Assume two classes with 0-1 encoding
+  c = 2;
+  class2 = find(T > 0.5);
+  TL = ones(n, 1);
+  TL(class2) = 2;
+  class2 = find(Y > 0.5);
+  Yclass = ones(n, 1);
+  Yclass(class2) = 2;
+end
+
+% Compute 
+correct = (Yclass==TL);
+total=sum(sum(correct));
+rate=[total*100/n total];
+
+C=zeros(c,c);
+for i=1:c
+  for j=1:c
+    C(i,j) = sum((Yclass==j).*(TL==i));
+  end
+end