Feldman and Friston 2010 "Attention, uncertainty, and free-energy" seem to include normally distributed noise in every generative function at every level; see e.g. p. 6, (14), which is supposed to be an expansion of (9).
Should I include Gaussian noise in generative functions? This wouldn't require changing levels.cljc, since the Gaussian could be added into the generative functions in the parameter files.
(One could cause the noise to be automatically added in what's done in levels.cljc, perhaps under control of a new field in Level. That would make the example files simpler, but would reduce the flexibility of the system. You'd have to have some way of running exercise 3 from Bogacz, for example.)
Feldman and Friston 2010 "Attention, uncertainty, and free-energy" seem to include normally distributed noise in every generative function at every level; see e.g. p. 6, (14), which is supposed to be an expansion of (9).
Should I include Gaussian noise in generative functions? This wouldn't require changing levels.cljc, since the Gaussian could be added into the generative functions in the parameter files.
(One could cause the noise to be automatically added in what's done in levels.cljc, perhaps under control of a new field in
Level. That would make the example files simpler, but would reduce the flexibility of the system. You'd have to have some way of running exercise 3 from Bogacz, for example.)