Normalization
Normalise by row, column or specified dimension.
Description
The Normalisation block independently normalises each row, column or vector of the specified input signal dimension. The output signal always has the same dimensions as the input signal.
This block treats input data of arbitrary dimension U as a set of vectors oriented along the specified dimension. The block normalises these vectors either modulo or modulo squared.
For example, consider a three-dimensional matrix U(i,j,k) and suppose you want to normalise it along the second dimension. First, define a two-dimensional intermediate quantity V(i,k) such that each element of V is the norm of one of the vectors in U :
Given V, the output of the block in 2-norm mode is:
In Squared 2-norm mode, the block output is:
Normalisation offset b is usually chosen as a small positive constant (e.g. 1e-10) which prevents potential division by zero.
Parameters
Norm - type of normalisation
2-norm (by default) | Squared 2-norm.
Normalisation type: 2-norm (modulo) or Squared 2-norm (modulo squared).
Normalisation bias - normalisation bias
1e-10 (By default)
The actual value of b to add to the denominator to avoid division by zero.
Normalise over - measurement to normalize
Specifies the dimension to normalise: rows, columns or the dimension specified in the Dimension parameter.
Dimension - dimension
1 (By default).
The measurement value, counted from the unit, by which normalisation is to be performed. The value of this parameter cannot exceed the number of measurements in the input signal.
This parameter is used when Normalise over is set to Specified dimension.