We shall now collect together the results derived in this section.
Substituting into equation 7 yields equation 15 for the expected description length LE.
Equations 13 and 14 describe how to convert a standard feedforward network into a network where the output of each neuron is assigned a mean and a variance. Equation 15 then defines the MDL-based cost function for such a network. The gradient of the cost function can be computed using standard backpropagation algorithm.