ctrl-z-9000-times
commented
2 years ago Yes, I think we definitely can do much better than these low scores...
I think some of the parameters got changed at some point; I don't think they match what numenta was using originally.
IIRC, Original Parameters are here: https://github.com/numenta/nupic/blob/master/src/nupic/frameworks/opf/common_models/anomaly_params_random_encoder/best_single_metric_anomaly_params_cpp.json
Here is the output of my new nyc_taxi.py program:
The TM is clearly not working. The blue line is the raw anomaly, which is almost always at 100%
As you can see, the scores went the wrong way.
I think part of the issue is the parameters for HTM-core are not great.