CMT

0.6.0

Added limited MATLAB interface
Improved Python 3 compatibility
Bug fixes

0.5.0

Regularizers are now more flexible
Several bug fixes and stability improvements
Added binomial distribution
Added StackedAffineTransform

0.4.1

Added multinomial logistic regression (MLR) - Added function for sampling images conditioned on component labels - Extended MCGSM by additional parameters

0.4.0

- Added spike-triggered mixture model (STM)

Added simple univariate distributions such as Bernoulli and Poisson
Added generalized linear model (GLM) and fully-visible belief network (FVBN)
Added mixture of Gaussian scale mixture (MoGSM)
Added PatchMCGSM
Extended MCGSM by linear features and means
Made implementation of new conditional models easier by introducing interface Trainable
Most methods of MCGSM can now cope with zero-dimensional inputs

0.3.0

Implemented early stopping based on validation error

0.2.0

Implemented mixture of conditional Boltzmann machines (MCBM)
Implemented PatchMCBM
Extended tools for generating data from images and videos