Changed implementation of our conditional MI calculators Kraskov via algorithm 2 to my current interpretation of how this should be done.
Added a utility for lagged covariance in MatrixUtils
Merged usual and debug methods for computing average MI in discrete calculator
Added setDebug to super InfoMeasureCalculator
Cleaned up header of InfoMeasureCalculator
Adding TODO comments for Kraskov calculators
Bug fix on MutualInfoCalculatorMultiVariateWithDiscreteKraskov - we weren't normalising incoming observations for computing local MI properly, this is fixed now.
Utilising determinant computation via Cholesky decomposition in Gaussian entropy and MI calculators
Allowed ChannelCalculatorCommon.finaliseObservations() to throw Exceptions so that child classes can do so (in particular Gaussian MI needs to do so if it finds the variables are linearly dependent)
Added routines in MathsUtils for uni- and multi-variate normal PDF and CDF (univariate only).
Implemented local MI for Gaussian MI. Pulled many routines into the MutualInfoMultiVariateCommon
Restructured MeasurementDistribution to have child classes EmpiricalMeasurementDistribution and AnalyticMeasurementDistribution - this resulted in many classes being altered.
Added ChiSquare distribution methods to MathsUtils.
Added more covariance methods to MatrixUtils.
Changed method signatures for initialisers for multivariate channel calculator (transfer entropy and mutual information) classes to have source dimensions before destination dimensions, in line with the addObservations (etc.) methods.
joseph.lizier