Derivative-free neural network for optimizing the scoring functions associated with dynamic programming of pairwise-profile alignment

Schematic diagram of the learning network. Upper case letters in italics and bold, lowercase letters in italics and bold, and lowercase letters in italics represent matrix, vector, and scalar values, respectively. Here, xa and xb represent the input vector, W1a, W1b, and w2 are weight matrices and vectors, b1 and b2 are bias vectors and scalar values, u is the middle layer vector, and y is the output value (the similarity score between PSSV A and PSSV B). The activating function is represented by φ(u). The square bracket represents the index of each vecto