A vector containing for each dimension
the number of units.
RETURNS
TRUE if an error was detected, FALSE if not.
FUNCTION
This function creates all the UNITs necessary to construct a Kohonen network.
In more detail: the double linked list of created units is attached to net through the
LastUnit and FirstUnit pointers. For each unit, the UnitName field is
filled, as are the UnitId and LayerId fields. The location (coordinates in
the line, matrix, cube or hypercube) are also entered in the UnitName field.
For each unit, the InOutIndex is calculated. Also, the biases (theta's) are
disabled by setting Theta.WeightFlag = FIXWEIGHT (see WEIGHT-flags and
Theta.Value.Weight = 1.0.
Kohonen networks always consist of two layers. The first is an input layer,
containing as many units as the dimensionality of the input. The second
layer contains the actual Kohonen map. This map can be 1, 2, 3 or 4 dimensional.
In any case, the input units do not use the transfer or activation functions;
i.e., they are set to NULL. The units in the Kohonen map get an
ActSqEucDist activation function and a TransLinear transfer
fucntion. All units in the input layer are fully connected to all units
in the map.
NOTE
This function is called by the create_koh_net function.
The malloc_unit function is used to allocate the unit structures.