Skip to main content
. Author manuscript; available in PMC: 2017 Jun 6.
Published in final edited form as: Neuroimage. 2015 Oct 23;125:498–514. doi: 10.1016/j.neuroimage.2015.10.045

Figure 1.

Figure 1

An illustration that shows heterogeneity in the discriminating boundary, relative to a reference group. Non-linear classifiers can implicitly model non-linearity but estimating the boundary (dashed lines) in high dimensional spaces is difficult. The proposed Mixture-of-Experts can approximate the non-linear curve with a piece-wise linear boundary (red and green lines) and find subgroups associated with each line (red and green points).