Daniel@0: Daniel@0:
Daniel@0:Daniel@0: mixparams = mdnfwd(net, x) Daniel@0: [mixparams, y, z] = mdnfwd(net, x) Daniel@0: [mixparams, y, z, a] = mdnfwd(net, x) Daniel@0:Daniel@0: Daniel@0: Daniel@0:
mixparams = mdnfwd(net, x)
takes a mixture density network data
Daniel@0: structure net
and a matrix x
of input vectors, and forward
Daniel@0: propagates the inputs through the network to generate a structure
Daniel@0: mixparams
which contains the parameters of several mixture models.
Daniel@0: Each row of x
represents
Daniel@0: one input vector and the corresponding row of the matrices in mixparams
Daniel@0: represents the parameters of a mixture model for the conditional probability
Daniel@0: of target vectors given the input vector. This is not represented as an array
Daniel@0: of gmm
structures to improve the efficiency of MDN training.
Daniel@0:
Daniel@0: The fields in mixparams
are
Daniel@0:
Daniel@0: Daniel@0: type = 'mdnmixes' Daniel@0: ncentres = number of mixture components Daniel@0: dimtarget = dimension of target space Daniel@0: mixcoeffs = mixing coefficients Daniel@0: centres = means of Gaussians: stored as one row per pattern Daniel@0: covars = covariances of Gaussians Daniel@0: nparams = number of parameters Daniel@0:Daniel@0: Daniel@0: Daniel@0:
[mixparams, y, z] = mdnfwd(net, x)
also generates a matrix y
of
Daniel@0: the outputs of the MLP and a matrix z
of the hidden
Daniel@0: unit activations where each row corresponds to one pattern.
Daniel@0:
Daniel@0:
[mixparams, y, z, a] = mlpfwd(net, x)
also returns a matrix a
Daniel@0: giving the summed inputs to each output unit, where each row
Daniel@0: corresponds to one pattern.
Daniel@0:
Daniel@0:
mdn
, mdn2gmm
, mdnerr
, mdngrad
, mlpfwd
Copyright (c) Ian T Nabney (1996-9) Daniel@0:
David J Evans (1998) Daniel@0: Daniel@0: Daniel@0: