Matlab Tools Documentation

Type “HELP FILE_NAME” for all details.

Filename Status

kmer_gapped_kernel.m

Computes the k-mer gapped kernel between the strings stored in S;
w gaps are allowed. The features are weighted by an
exponential lambda^-m, where m is the actual number of gaps for that feature.

final

kmer_kernel.m
Computes the k-mer kernel between the strings stored in S

final

kmer_wildcard_kernel.m
Computes the k-mer wildcard kernel between the strings stored in S; w wildcards are allowed. The features are weighted by an exponential lambda^-m, where m is the actual number of wildcards for that feature.

final
bow_kernel.m
This function computes the bag of words kernel matrix for the strings in S. The case of the letters is ignored.
final
bag_of_words.m
function kernel=bag_of_words(s,t)
This function computes the bag of words kernel between two strings
(not optimal, just as example)
centering.m
Centers the matrix K
final

cholesky.m
Performs incomplete cholesky of the kernel matrix K

 

dualcca.m
Performs (multiway) CCA between the kernels in the different cells of Ks.

 

dualfisher.m
Dual (kernel) Fisher discriminant analysis (KFDA).

final

dualkmeans.m
Performs dual K-means for ell samples specified by the kernel K

 

dualpca.m
Performs dual/kernel PCA

 

dualpls.m
Performs dual (kernel) PLS discrimination

mmm

normalise.m
Normalises the kernel matrix

final

pls.m
Performs PLS discrimination

 

simplenovelty.m
simple novelty detection algorithm based on distance from center of mass

final

standardise.m
standardizes data in feature space

 

visualise.m
visualization algorithm

 
diag_red.m
Based on Kandola et al SDP methods for diagonal reduction
requires SEDUMI
matlab
toolbox
rbf.m  
spectral_clustering.m

implementation of Ng, Weiss, Jordan methods

 
svmnovelty.m
Performs novelty detection based on a sample stored in K, for test samples specified in Ktest. Note that this is a naive version, simply making use of matlab’s built-in qp-solver.
needs optimization toolbox
svmctrain.m
Computes the dual vector and the offset for support vector machine classification.
Note that this is a naive version, simply making use of matlab’s built-in qp-solver.
requires optim toolbox
adjrand.m
measure of “similarity” between two clusterings of the data
final
anova.m  
p_spectrum.m

computes the p-spectrum kernel (dynamic programming) outputs the DP table

purely for teaching purposes
p_spectrum_bf.m
(computes the p-spectrum kernel by brute force)
purely for teaching purposes
p_spectrum_fast.m
(dynamic programming faster implementation of p spectrum kernel) outputs the DP table
purely for teaching purposes
p_trie.m
(p spectrum kernel again, based on tries)
purely for teaching purposes
blended_spectrum.m
(blended spectrum up to p by dyn prog) outputs the DP table
purely for teaching purposes
blended_spectrum_bf.m
(blended spectrum up to p by brute force)
purely for teaching purposes
blended_spectrum_fast.m
(blended spectrum up to p – faster) outputs the DP table
purely for teaching purposes
subseq_count .m
(same as discussed in book, for non continuous subsequences) outputs the DP table
purely for teaching purposes
 ssk_fast.m
(as in ch11, outputs the DP table)
purely for teaching purposes

WARNING: Do not use this software to control an aircraft or laser eye surgery. We cannot guarantee it is fully debugged.