Using the principles we explored in our meeting (Jan. 20), we can add some form of memory into the model using recurrence. This can be easily integrated into our existing MELM as follows:
with nengo.Network() as model:
# Existing MELM ensemble and connection construction
rep_recur_con = nengo.Connection(rep_ens, rep_ens, "transform", "synapse")
# Existing MELM probingWe can also consider a separately integrating and block that contributes to the output just like the representation ensemble:
(I apologize, I drew this with my computer mouse and it is... not incredible)
Just like with the MELM, we're interested in seeing how this model performs when trained and tested on our spiking dataset. Since we're concerned with performance over time, plotting the error (absolute and squared) over time could be a useful exercise in visualizing where the model struggles.
Similarly, it would be useful to generate visualizations of the RLM model using nengo-gui for a more visual and intuitive way to interpret the underlying architecture (my drawing is not adequate).
Using the principles we explored in our meeting (Jan. 20), we can add some form of memory into the model using recurrence. This can be easily integrated into our existing MELM as follows:
We can also consider a separately integrating and block that contributes to the output just like the representation ensemble:
(I apologize, I drew this with my computer mouse and it is... not incredible)
Just like with the MELM, we're interested in seeing how this model performs when trained and tested on our spiking dataset. Since we're concerned with performance over time, plotting the error (absolute and squared) over time could be a useful exercise in visualizing where the model struggles.
Similarly, it would be useful to generate visualizations of the RLM model using nengo-gui for a more visual and intuitive way to interpret the underlying architecture (my drawing is not adequate).