ReconstructionICA
statistics: ReconstructionICA
Reconstruction independent component analysis (RICA) feature-extraction model.
A ReconstructionICA object stores the transformation learned by
rica for extracting features from data. Create one with rica;
apply it to data with the transform method.
See also: rica, sparsefilt
Source Code: ReconstructionICA
The ReconstructionICA class contains the following properties:
A scalar structure holding the options the fit ran with:
IterationLimit, Lambda, Standardize,
ContrastFcn, InitialTransformWeights,
GradientTolerance, StepTolerance, Solver and
NonGaussianityIndicator.
This property is read-only.
A positive integer P, the number of columns of the training data. This property is read-only.
A positive integer Q, the number of features the learned transformation produces. This property is read-only.
A column vector with one entry per predictor, the mean of each
column of the training data. It is empty unless 'Standardize'
was true. This property is read-only.
A column vector with one entry per predictor, the standard deviation
of each column of the training data. It is empty unless
'Standardize' was true. This property is read-only.
A scalar structure with the fields Iteration and
Objective, both column vectors of the same length.
Iteration counts from zero and Objective(1) is the
objective at the starting weights, so the last entry of each is the
solution the fit returned. This property is read-only.
The trajectory is this implementation’s own. The default
'quasinewton' solver minimises through Octave’s fminunc,
and 'Solver', 'lbfgs' selects the limited-memory BFGS solver
MATLAB uses. Either way the steps taken from the same starting weights
differ from MATLAB’s, so the length of the history and the iteration
counts differ, and on an objective this far from convex the optimum
reached need not be MATLAB’s either.
A P-by-Q matrix of learned weights, its columns of unit
length. The transform method applies it to data. This property
is read-only.
A P-by-Q matrix, the weights the fit started from. It
is the matrix given as 'InitialTransformWeights' when one was
given, and the random start the fit drew otherwise. This property is
read-only.
A Q-by-1 vector of +1 and -1, one per learned feature: +1 where the feature is taken to be super-Gaussian and -1 where it is taken to be sub-Gaussian. The entry sets the sign its feature’s contrast term carries in the objective, so the fit seeks a sparse feature where the entry is +1 and a spread one where it is -1. The default is all +1. This property is read-only.
The ReconstructionICA class offers the following public methods:
statistics: Mdl = ReconstructionICA (X, Q, …)
statistics: Z = transform (Mdl, X)