Listing 1. Pseudo-code for the C/C++ template used for a DNN layer.

// Multiply the bias value by the weight of j-th neuron for j = 1 to N_Output  B[j] = Wb[j] ∗ b end // Multiply and accumulate layer weights for all neurons for i = 1 to N_Input  for j = 1 to N_Output   X[j] = X[j] + x[i] ∗ W[j]  end end // Add weighted bias and weighted neurons and produce output neurons // result via activation function for j = 1 to N_Output  y[j] = f(X[j]+B[i])  // activation function of each neuron at k-th layer end