Hi Marcus, I spent some time playing with this project, and I have to say that it is very well-made. Hats off to you and Felix for creating such a polished and well-thought out visualization tool.
I like that you can look at what's going on in the TM region in great detail. It's both fun and satisfying to look at the different cells and columns and see what they are connected to. When debugging a research project I think it would be really useful to also have higher-level statistics (for example, # of predicted cells that then became active, graphed over time). Take a look at the TemporalMemoryMonitorMixin [1] in nupic; this is what we use often in research projects. If this visualization project also included those kinds of high level graphs, it would help provide a more complete picture of what's going on. That being said, I can see this being potentially useful especially in educational settings; for example, presentations that describe the temporal memory. It gives a really clear picture of how columns and cells are connected, and why cells are becoming predicted and active. A nice little project for an interested community member might be to create an example of high-order sequence learning, visualize it with Sanity, and make a screencast or an interactive document that demonstrates how the temporal memory learns high order sequences. Hope this feedback is helpful. Again, you and Felix have done a wonderful job on this project, thanks for the contribution. [1] https://github.com/numenta/nupic/blob/master/src/nupic/research/monitor_mixin/temporal_memory_monitor_mixin.py - Chetan > On Nov 23, 2015, at 5:13 PM, Marcus Lewis <[email protected]> wrote: > > Hi all, > > I made a sequel: See your HTM run: Duct tape is okay! > <https://www.youtube.com/watch?v=bqu-hc4pc7Q> (3:17) > > <https://www.youtube.com/watch?v=bqu-hc4pc7Q> > <image.png> <https://www.youtube.com/watch?v=bqu-hc4pc7Q> > > I ported a nupic.research feedback experiment to this environment, and I use > it to show off apical dendrites. I also ported a Union Temporal Pooler > experiment, which you can try > <https://github.com/nupic-community/sanity-nupic/blob/master/examples/research_union_pooler.py>, > though it's not in the video. > > My goal is to demonstrate that this kind of thing is practical for HTM > research. > > Hope you like it! > Marcus
