Principal Feature Analysis: A Multivariate Feature Selection Method for fMRI Data

Figure 3

Similarity matrices for stimuli of four categories. The correlations between each pair of patterns corresponding to 200 stimuli (50 stimuli for each category) were calculated. These patterns were normalized to a mean of zero in each voxel across categories by subtracting the mean response across all patterns. The color reflects the values of the correlations (see color bar), patterns with high positive correlations would be classified to the same category, whereas those with low negative correlations would be classified to different categories.