run rsa visualize

Matlab output: run_rsa_visualize

%% RSA Visualize
%
% #   For CoSMoMVPA's copyright information and license terms,   #
% #   see the COPYING file distributed with CoSMoMVPA.           #

%% Load data in EV and VT mask
% load datasets cosmo_fmri_dataset

config=cosmo_config();
data_path=fullfile(config.tutorial_data_path,'ak6','s01');

data_fn=[data_path '/glm_T_stats_perrun.nii'];
targets=repmat(1:6,1,10)';
ev_ds = cosmo_fmri_dataset(data_fn, ...
                            'mask',[data_path '/ev_mask.nii'],...
                            'targets',targets);

vt_ds = cosmo_fmri_dataset(data_fn, ...
                            'mask',[data_path '/vt_mask.nii'],...
                            'targets',targets);

% compute average for each unique target, so that the datasets have 6
% samples each - one for each target
vt_ds=cosmo_fx(vt_ds, @(x)mean(x,1), 'targets', 1);
ev_ds=cosmo_fx(ev_ds, @(x)mean(x,1), 'targets', 1);


% remove constant features
vt_ds=cosmo_remove_useless_data(vt_ds);
ev_ds=cosmo_remove_useless_data(ev_ds);

% Use pdist (or cosmo_pdist) with 'correlation' distance to get DSMs
% in vector form. Assign the result to 'ev_dsm' and 'vt_dsm'
ev_dsm = pdist(ev_ds.samples, 'correlation');
vt_dsm = pdist(vt_ds.samples, 'correlation');

model_path=fullfile(config.tutorial_data_path,'ak6','models');
load(fullfile(model_path,'behav_sim.mat'));
behav_dsm=squareform(behav);


% Using matlab's subplot function place the heat maps for EV, VT
% and behaviour DSMs side by side in the top three positions of a 3 x 3
% subplot figure.
% (Hint: to convert DSMs in vector form to matrix form (and vice versa),
% using cosmo_squareform or squareform).

figure();

subplot(3,3,1);
imagesc(squareform(ev_dsm));
title('EV');

subplot(3,3,2);
imagesc(squareform(vt_dsm));
title('VT');

subplot(3,3,3);
imagesc(squareform(behav_dsm));
title('behav');

labels = {'monkey','lemur','mallard','warbler','ladybug','lunamoth'}';

%% Add the dendrograms for EV, LV and behav in the middle row of the
% subplot figure (this requires matlab's stats toolbox)
if cosmo_check_external('@stats',false)
    % First, compute the linkage using Matlab's linkage for
    % 'ev_dsm', 'vt_dsm' and 'behav_dsm'. Assign the result
    % to 'ev_hclus', 'vt_hclus', and 'behav_hclus'

    ev_hclus = linkage(ev_dsm);
    vt_hclus = linkage(vt_dsm);
    behav_hclus = linkage(behav_dsm);


    subplot(3,3,4);
    % show dendrogram of 'ev_hclus'
    % As additional arguments to the dendrogram function, use:
    %      'labels',labels,'orientation','left'
    dendrogram(ev_hclus,'labels',labels,'orientation','left');

    % Using the same approach, show a dendrogram of 'vt_hclus'
    subplot(3,3,5);
    dendrogram(vt_hclus,'labels',labels,'orientation','left');

    % Using the same approach, show a dendrogram of 'behav_hclus'
    subplot(3,3,6);
    dendrogram(behav_hclus,'labels',labels,'orientation','left');
else
    fprintf('stats toolbox not present; cannot show dendrograms\n');
end

%% Show the MDS (multi-dimensional scaling) plots in the bottom row

% Show early visual cortex model similarity
subplot(3,3,7);

% get two-dimensional projection of 'ev_dsm' dissimilarity using cmdscale;
% assign the result to a variable 'xy_ev'
xy_ev = cmdscale(squareform(ev_dsm));

% plot the labels using the xy_ev labels
text(xy_ev(:,1), xy_ev(:,2), labels);

% adjust range of x and y axes
mx = max(abs(xy_ev(:)));
xlim([-mx mx]);
ylim([-mx mx]);

% Show VT similarity

% using cmdscale, store two-dimensional projection of 'vt_dsm' and
% 'behav_dsm' in 'xy_vt' and 'xy_behav'
xy_vt = cmdscale(squareform(vt_dsm));
xy_behav = cmdscale(squareform(behav_dsm));

subplot(3,3,8);
text(xy_vt(:,1), xy_vt(:,2), labels);
mx = max(abs(xy_vt(:)));
xlim([-mx mx]);
ylim([-mx mx]);


subplot(3,3,9);
text(xy_behav(:,1), xy_behav(:,2), labels);
mx = max(abs(xy_behav(:)));
xlim([-mx mx]);
ylim([-mx mx]);