Deep Learning Toolbox Data Conventions

Dimensions

The following code dimensions are used in describing both the network signals that users commonly see, and those used by the utility functions:

Ni = Number of network inputs	= net.numInputs
Ri = Number of elements in input i	= net.inputs{i}.size
Nl = Number of layers	= net.numLayers
Si = Number of neurons in layer i	= net.layers{i}.size
Nt = Number of targets
Vi = Number of elements in target i, equal to Sj, where j is the ith layer with a target. (A layer n has a target if net.targets(n) == 1.)
No = Number of network outputs
Ui = Number of elements in output i, equal to Sj, where j is the ith layer with an output (A layer n has an output if net.outputs(n) == 1.)
ID = Number of input delays	= net.numInputDelays
LD = Number of layer delays	= net.numLayerDelays
TS = Number of time steps
Q = Number of concurrent vectors or sequences

Variables

The variables a user commonly uses when defining a simulation or training session are

P	Network inputs	Ni-by-TS cell array, where each element P{i,ts} is an Ri-by-Q matrix
Pi	Initial input delay conditions	Ni-by-ID cell array, where each element Pi{i,k} is an Ri-by-Q matrix
Ai	Initial layer delay conditions	Nl-by-LD cell array, where each element Ai{i,k} is an Si-by-Q matrix
T	Network targets	Nt-by-TS cell array, where each element P{i,ts} is a Vi-by-Q matrix

These variables are returned by simulation and training calls:

Y	Network outputs	No-by-TS cell array, where each element Y{i,ts} is a Ui-by-Q matrix
E	Network errors	Nt-by-TS cell array, where each element P{i,ts} is a Vi-by-Q matrix
perf	Network performance

Utility Function Variables

These variables are used only by the utility functions.

Pc	Combined inputs	Ni-by-(ID+TS) cell array, where each element P{i,ts} is an Ri-by-Q matrix Pc = [Pi P] = Initial input delay conditions and network inputs
Pd	Delayed inputs	Ni-by-Nj-by-TS cell array, where each element Pd{i,j,ts} is an (Ri*IWD(i,j))-by-Q matrix, and where IWD(i,j) is the number of delay taps associated with the input weight to layer i from input j Equivalently, IWD(i,j) = length(net.inputWeights{i,j}.delays) Pd is the result of passing the elements of P through each input weight's tap delay lines. Because inputs are always transformed by input delays in the same way, it saves time to do that operation only once instead of for every training step.
BZ	Concurrent bias vectors	Nl-by-1 cell array, where each element BZ{i} is an Si-by-Q matrix Each matrix is simply Q copies of the net.b{i} bias vector.
IWZ	Weighted inputs	Ni-by-Nl-by-TS cell array, where each element IWZ{i,j,ts} is an Si-by-???-by-Q matrix
LWZ	Weighted layer outputs	Ni-by-Nl-by-TS cell array, where each element LWZ{i,j,ts} is an Si-by-Q matrix
N	Net inputs	Ni-by-TS cell array, where each element N{i,ts} is an Si-by-Q matrix
A	Layer outputs	Nl-by-TS cell array, where each element A{i,ts} is an Si-by-Q matrix
Ac	Combined layer outputs	Nl-by-(LD+TS) cell array, where each element A{i,ts} is an Si-by-Q matrix Ac = [Ai A] = Initial layer delay conditions and layer outputs.
Tl	Layer targets	Nl-by-TS cell array, where each element Tl{i,ts} is an Si-by-Q matrix Tl contains empty matrices [] in rows of layers i not associated with targets, indicated by net.targets(i) == 0.
El	Layer errors	Nl-by-TS cell array, where each element El{i,ts} is an Si-by-Q matrix El contains empty matrices [] in rows of layers i not associated with targets, indicated by net.targets(i) == 0.
X	Column vector of all weight and bias values