[Nengo-DL]: Adding a dynamic custom layer between two Ensembles

[zerone]

Hello @xchoo, NP and thanks for getting back.

With respect to the following:

@xchoo

you are correct in the observation that you do not see any filtered spikes. This is because the filtering is done on the output of a connection, which is then accumulated on the input of the Ensemble / Node.

you previously in another post (I don’t remember which) noted that filtering is done while inputing to the Ensembles (and not at the output of the Ensemble), however, I guess, this statement is equivalent to

@xchoo

the filtering is done on the output of a connection

i.e. output of a connection is exactly what’s input to the Ensemble/Node (with no other in-between processing of values).

With respect to the following,

@xchoo

you’ll need to probe the connection that connects to that node

I haven’t tested it yet, but just confirming with you… it still won’t work for the TensorNode in question right? as you realize it as a bug in current Nengo-DL version… isn’t it?

However, with respect to the bug you raised, if one still wishes to work on the spike amplitude output (and not the filtered one), then I guess he/she has to explicitly set the synapse of Connection to the respective Node/Ensemble/TensorNode as None. Isn’t it?

With respect to the following,

@xchoo

This is indeed the case in the code.

it means that ndl_model.layers and ndl_model.net.ensembles are essentially representing the same thing… right?

Yeh , thanks for mentioning the following… I should try it as well.

@xchoo

some kind of averaging or smoothing to the spiking output

With respect to the following,

@xchoo

It would be the same for all layers, and by that I mean that the filtered spikes is used as an input to each layer

just explicitly confirming, the filtered “spike train” is considered as x at all the layers and current J is calculated for each of them. Isn’t it?