Hi I’m using HTM for anomaly detection. Is there a possibility to get or print a probability measure for an detected anomaly in real time. e.g this detected anomaly get the probability of 0.98 to be a real anomaly. I want to do a kind of soft clustering like in Gaussian Mixture Models. thanks in …

Use the anomaly likelihood, which is a logarithmic number, so set an anomaly threshold very high, like 0.99999.

Probability of detected anomaly to measure uncertainty

learner August 29, 2018, 12:50pm 2

Using the Anomaly Likelihood doesn’t meet your specs?

Topic		Replies	Views
Please throw me a clue: finding anomaly _likelihood_ HTM.Java	2	1310	May 19, 2017
Combining multiple metrics into a likelihood Engineering question	7	620	April 23, 2020
Clarification on anomaly score and anomalylikelihood Numenta Theory	14	4096	October 12, 2019
Getting anomaly likelihood NuPIC	7	593	January 29, 2020
Understanding AnomalyLikelihood? Engineering	1	415	October 8, 2019