The base graph object, FFNeuralGraph encapsulates an existing graph object of type FramedGraph and builds upon it by defining a set of behaviours expected from Neural Network graphs
A list of Strings containing the activations for each layer. Options for activation functions are: 1) "logsig" or "sigmoid" 2) "tansig" 3) "linear" 4) "recLinear"
The number of hidden layers in the network.
Perform a forward pass through the network to calculate the predicted output.
Perform a forward pass through the network to calculate the predicted output.
Get as a scala Iterable the neurons for a particular layer.
Get as a scala Iterable the neurons for a particular layer.
The layer number, can vary from 0 (input layer) to hidden_layers + 1 (output layer)
The neurons in the particular layer as Neuron objects
Get as a scala Iterable the synapses between layer l and l-1.
Get as a scala Iterable the synapses between layer l and l-1.
The layer number, can vary from 1 (input layer synapses) to hidden_layers + 1 (output layer synapses)
The respective synapses as Synapse objects
Get as a breeze DenseMatrix the synapses between layer l and l-1.
Get as a breeze DenseMatrix the synapses between layer l and l-1.
The layer number, can vary from 1 (input layer synapses) to hidden_layers + 1 (output layer synapses)
The respective synapses as elements of a matrix
Perform a forward pass through the network to calculate the predicted output for a batch of input points.
Perform a forward pass through the network to calculate the predicted output for a batch of input points.
The input batch as a List of Lists where each level of the top level List represents an input node. On the other hand each element of the lower level list represents a particular dimension of a particular data point in the data set.
Represents the underlying graph of a feed-forward neural network.