Figure 3
From: Learning sparse models for a dynamic Bayesian network classifier of protein secondary structure
Schematic of the dlink structure, used to represent correlations in the DBN. Each node represent a PSSM element and edges represent the dlink coefficients of the DBN. The rightmost column corresponds to the ith vector of the PSSM Xi (i.e., the vector for the ith position along the amino acid sequence). The remaining columns represent the columns of the PSSM that come before the ith position if we traverse the sequence from the N-terminal to the C-terminal of the protein. Rows represent the 20 amino acids in the PSSM. For simplicity only edges for the child node x i (1) is shown, which is the first element of the observation vector at the ith position. The child nodes are in the rightmost column and the parent nodes are in positions i - L AA , ..., i.