↑ Return to Neural networks

Mean field activity

In this case a layer of interconnected leaky neurons is characterised by a positive tanh as a transfer function. The time component is slower than the one we have in spiking neurons (the decay of the leaky is at least 100 times slower), as a consequence each of these slow units can be conceived as simulating the average activity of a whole neural population or cluster of spiking neurons.

These type of units have been used, for instance, here: http://journal.frontiersin.org/article/10.3389/fpsyg.2014.00124/full


NB the simulation require the use of a step function, see below.


Step function to update the activity of the units in the cluster. Save this code as a separate file with the name “leaky_step” in the same folder as the main file above.


Leave a Reply

%d bloggers like this: