eminorhan / inevitable-probability

Efficient probabilistic inference in generic neural networks trained with non-probabilistic feedback

Inevitable Probability

This repository contains the code and data used to obtain the results reported in the following paper:

Orhan AE, Ma WJ (2017) Efficient probabilistic inference in generic neural networks trained with non-probabilistic feedback. Nature Communications, 8, 138.

The bulk of the code is written in Theano (0.8.2) + Lasagne (0.2.dev1). Each folder contains code and data pertaining to a particular model type or experiment:

• alt_objectives: training with alternative objectives
• ffwd: experiments with feedforward nets
• nin_nhu: experiments measuring the efficiency of generic nets
• random_ffwd: experiments with random feedforward nets
• recurrent_ei: experiments with recurrent excitatory-inhibitory (EI) nets

The code for generating the results reported in Figure 9 (qamar2013) is written in Matlab and uses some routines from the Matlab Statistics and Optimization Toolboxes. Some of the files are meant to be run on a local computer cluster. You may need to modify them according to your needs.