circreg circular regression method.
DESCRIPTION
Circular regression using the Fisher and Lee model.
EXAMPLE
% generate data using fixed mean mu and concentration kappa
X = randn(1000,3);
c = dml.circreg('mu',0,'kappa',100);
Y = c.sample(X);
ix = (-pi:0.1:pi)';
x = histc(Y,ix); x = x ./ sum(x);
polar(ix,1+x);
% check estimation
c = dml.circreg('mode','none','verbose',true);
c = c.train([],Y);
disp([c.mu c.kappa]);
% generate data using X-dependent mean mu and concentration kappa
X = randn(1000,3);
c = dml.circreg('mu',0,'beta',[-1 0 1]','kappa',100);
Y = c.sample(X);
% check estimation
c = dml.circreg('mode','mean','verbose',true);
c = c.train(X,Y);
disp([c.mu c.kappa c.beta']);
DEVELOPER
Marcel van Gerven (m.vangerven@donders.ru.nl)
Ali Bahramisharfi (ali@cs.ru.nl)
| beta |
regression coefficients for the mean |
| gamma |
regression coefficients for the concentration |
| indims |
dimensions of the input data (excluding the trial dim and time dim in time series data) |
| inner |
maximum number of inner loop iterations in concentration/mixed estimation |
| kappa |
concentration parameter of von Mises distribution |
| lambda |
regularization parameter for the mean (only defined for gradient descent) |
| likelihood |
stored loglikelihood |
| method |
method used to estimate model (1=standard, 2=generalized method of moments, 3=second harmonics) |
| mu |
intercept of von Mises distribution |
| outer |
maximum number of outer loop iterations in mixed estimation |
| repeat |
number of repeats for gradient descent (multiple local maxima) |
| restart |
when false, starts at the previously learned parameters; needed for online learning and grid search |
| tol |
smallest update step in concentration/mixed estimation |
| verbose |
whether or not to generate diagnostic output |