Now that we've cleaned up our data, created a training or test set, and written the code necessary to produce the input required by our network, we can begin coding up the RBM itself.
First, we begin with our now standard struct, the scaffold onto which we attach the various components of our network:
const cdS = 1type ggRBM struct { g *ExprGraph v *Node // visible units vB *Node // visible unit biases - same size tensor as v h *Node // hidden units hB *Node // hidden unit biases - same size tensor as h w *Node // connection weights cdSamples int // number of samples for contrastive divergence - WHAT ABOUT MOMENTUM}func (m *ggRBM) learnables() Nodes { return Nodes{m.w, m.vB, m.hB}}Then, we add the helper functions ...
