Returns an initial random weight matrix.
Generates a weight matrix with the bias weights set to map to the mean of the target distribution and remaining weights drawn at random from a Gaussian distribution with zero mean and variances choosen so that the variances of the generated distribution roughly match the variances of the target distribution.
In addition, an initial value for beta may be calculated as the inverse of the average distance between each Gaussian centre, calculated with the random mapping, and its nearest neighbours in the set of data points.
[W, beta] = gtm_ri(T, FI)
W = gtm_ri(T, FI)
T - sample of target distribution, used for calculating the mean; one data point per row; N-by-D
FI - basis functions' activations when fed the latent data, X, plus a bias, K-by-(M+1)
W - the initialised weight matrix; (M+1)-by-D
beta - the initial beta value, scalar. This is an optional output argument; if omitted, the corresponding
calculations are omitted too.