Welcome to IMNN’s documentation!

The information maximising neural network (IMNN) is a fitting algorithm for neural networks that aims to maximise the Fisher information of a training set to produce the most informative set of summaries about the model parameters of a generative model of some target data. In particular, an IMNN can extract, asymptotically losslessly, information from complex distributed data in a \(d\)-dimensional space and map it to some normally distributed summaries in a \(n_\rm{params}\)-dimensional space, where \(n_\rm{params}\) is the number of parameters in the generative model for the data.

doc pypi bit git doi zen

Indices and tables