HTM NUPIC - anomaly score - different result - same settings for SP,TM,encoder

Engineering
6

There are two things we should address.

  1. different results for different runs of NuPIC with the same data
  2. what you are trying to accomplish and how to accomplish it

Let’s tackle #1 first. I just want to clarify again. When you run your script and process a data set, can you output the results to a file? Then run the script again and output to another output file and diff the results. I want to see the diff. I do not expect for results to be different unless:

  • input data changed (even one bit)
  • random seed changed
  • model configuration changed
  • model being used was saved to disk between runs and re-used

We can talk about #2 once we get #1 resolved.

1 Like