Ben Morcos
Ben Morcos
Looking for some opinions on what is more intuitive. Let's say you want to apply a learning rule only **every 4 steps**, should you pass the parameter: - As 4...
Another discussion regarding the PES learning rule implementation in `nengo/builder/learning_rules.py`. Right now the PES rule does not implement it's own operator class like the other learning rules do (ex. `class...
Another point raised by @tbekolay, if we apply the learning rule only sometimes, ie. ``` period = (1 if self.apply_every is None else apply_every / self.dt) def step_simvoja_sometimes(): if n_steps...
Update after quick a discussion in the dev meeting: I am planning to add scaling to the `delta` signal to make the learning rate more stable between normal learning and...
I have code working that takes an argument `apply_every` in seconds, I then calculate the period (in terms of `dt` steps) at which to apply learning. If learning is not...
This is the basic script I've been playing with in the GUI: ``` import nengo from nengo.processes import WhiteSignal import numpy as np learning_rate = 1e-4 dt = 0.001 #...
When I was running Nengo on a small embedded platform I kept running out of memory. I am using ReLu so I made the change above and I no longer...
Would this be a good issue to tackle to try and ease into the GUI code?
Where would be a good place to start looking at code?
I finally got this to compile this morning. I found some related mbed issues/PRs, specifically this one: ARMmbed/mbed-os#13119, which led me to a fix. By default TFMicro downloaded mbed-os at...