Bearded-android-docs

Created Monday 28 April 2014

A neural network "learns" from training data or training examples. A single training example is an input vector `bb x` and the expected (or desired) output `bb t` at that input.

The "complexity" lies in the notation:

• we'll denote multiple training data by `{(bb x[1], bb t[1]), ...,(bb x[p], bb t[p])}` where `p` is the size of the training data. `(bb x[1], bb t[1])` is the first training example, `(bb x[2], bb t[2])` is the second training example, and so on...
• each input and expected output are vectors, `bb x[i] = (x[i]_1, ..., x[i]_n)` and `bb t[i] = (t[i]_1, ..., t[i]_m)` where `n` and `m` are the dimensions of the input and output spaces, respectively.
• Each `t[i]_j` is the expected output of the ith training example for the jth neuron in the output layer.