Permalink
Browse files

Update to cross_validation to reject singles in patterns, updates to …

…blur_pattern to generate doubles for patterns by default -GTM
  • Loading branch information...
Wildcarde committed Feb 8, 2011
1 parent 0f78793 commit bb4de22e6cdf59ec63ad9a2629af8146a765ea4f
Showing with 6 additions and 2 deletions.
  1. +2 −0 core/learn/cross_validation.m
  2. +4 −2 core/preproc/blur_pattern.m
@@ -220,6 +220,8 @@
cur_maskname = masknames{n};
masked_pats = get_masked_pattern(subj,cur_patname,cur_maskname);

assert(strcmp(class(masked_pats),'double'));

% Create the training patterns and targets
trainpats = masked_pats(:,train_idx);
traintargs = regressors( :,train_idx);
@@ -13,7 +13,7 @@
% Optional Arguments:
%
% 'single' - Whether or not to use single precision to save
% memory.
% memory. Default False.
%
% 'n' - The size of the gaussian filter in voxels. (Default: ceil(FHWM))
%
@@ -34,7 +34,7 @@
%
% ======================================================================

defaults.single = true;
defaults.single = false; %defautl changed 2-8-2011 GTM, responce to bug report by Zhen James Xiang on 1-25-2011
defaults.n = ceil(fwhm);
defaults.new_patname = sprintf('%s_sm%g', patname, fwhm);

@@ -70,6 +70,8 @@

if args.single
pat = single(pat);
else
pat = double(pat);
end

% Insert the new pattern

0 comments on commit bb4de22

Please sign in to comment.