Figure 1.

The hybrid Bayesian network. The probability of belonging to cluster c is estimated from the brain response z from a high-level brain area with the help of γ₀ and γ. An encoding model provides regression coefficients B and covariance Σ to reconstruct the image from the brain response y of a lower-level brain area. The mean m_c and covariance R_c of a set of prior images enforce structure and the cluster c enforces semantic information during the reconstruction of image x. This process can be repeated for a set of N images.