Hey there,

I’m trying to implement anomaly detection along the lines of the discussion in supplementary section S4 of the “Unsupervised real-time anomaly detection for streaming data” paper (Pages 6-8 of https://ars.els-cdn.com/content/image/1-s2.0-S0925231217309864-mmc1.pdf)

Let’s say I have two models, which output prediction errors `s1_t`

and `s2_t`

at every time, `t`

. The goal of the discussion is to be able to detect if the prediction error of the first model spikes at a different but close time to the second (i.e. `s1_4`

and `s2_5`

are spikes). I’m very confused why they propose including `G`

, a Gaussian convolution kernel, which seems to use `x`

, the input to HTM (the raw value).

Could someone walk me through the math of this section, and if possible, how one would implement it?

Thanks!