Notation used in the slides:
N - number of training points
xn - index of training point
Q - input dimensionality
D - output dimensionality
C - number of classes in classification
K - number of units (neurons), often in last layer
X - training inputs (N by Q matrix)
x - single training input (Q by 1 vector)
y - single training output (D by 1 vector)
D = {(x1, y1), ..,(xN, yN)} = X, Y - training set
W - neural network weight matrix (often last layer, often a random variable)
theta - variational parameters