Eric Hunsberger
Eric Hunsberger
The emulator and chip still differ on their learning. - [ ] `test_pes_comm_channel` has differences between emulator and chip, and these differences are not consistent between different numbers of neurons...
This isn't an issue, but a challenge! Points for anyone who can make compartment voltage overflow on the chip. Originally, I thought that V was 23 bits + sign. However,...
With #132, the oscillator actually works fairly well now. One remaining problem is that it oscillates more slowly than expected. My best explanation is that the interneurons that we need...
On some models (including some Nengo tests), our current code to remove passthrough nodes is failing. This is often because removing the passthrough node entirely turns a factored `(m, 1)...
Connections with interneurons (i.e. all decoded connections without weight solvers) have a lot of noise (they use 20 noisy interneurons to transmit information). This noise has a particularly adverse effect...
Currently, the amplitude test is not running (since ReLUs are not implemented). Furthermore, we need proper testing for when the pre or post part of a connection is off-chip. These...
Currently, the script console can't handle interactive stuff. One example of this that would be really nice to have would be interactive debugging with pdb. As an illustration, go to...
For a course project, I'm going to be looking at faster ways to solve for decoders in Nengo. Whatever method I decide on should be implemented by the end of...
If I add a an origin, choose user-defined function, and specify just a function as the expression (e.g. "min"), it breaks Nengo. There should be more checking that the expression...
If I create an ensemble and add an origin that computes a function (I was testing with a quadratic) and then save with scriptgen, this origin is not saved. Furthermore,...