search

fmi-basel / latent-predictive-learning

Code to accompany our paper "The combination of Hebbian and predictive plasticity learns invariant object representations in deep sensory networks” bioRxiv 2022
MIT License
21 stars 5 forks source link

Missing dt in sigma2 trace normalization? #4

Closed yilun-wu closed 8 months ago

yilun-wu commented 10 months ago

Dear authors,

I believe a multiplication of auryn_timestep is missing from the following normalization of the trace to be consistent with Eq. 20, can you confirm? https://github.com/fmi-basel/latent-predictive-learning/blob/dd486d092b8cea9bfa1f27f42209f9174b82e61c/spiking_simulations/LPLConnection.cpp#L485-L485

Of course, I understand the impact of this can be absorbed in $\lambda$ in Eq. 18.

fzenke commented 9 months ago

Looks like you are right. Thanks for flagging this!