[linvs]

As someone that works closely with WiFi data and given their very low error (< 0.5 BMP MAE) I'd love to see them address a few key points:

- Whether train/test splits were participant-wise to avoid data leakage.

- How the system performs at elevated or highly variable heart rates.

- Results from "placebo" or empty-room baselines to rule out false positives, typically done with bags of rice/water (used to simulate mass).

[PranayKo]

1. the train test in here was leaky, but in our current iteration of the work we do 10 fold train / test without leakage 2. There was no different in performance at higher HR values, the rpi data contained people running in place and our performance on that was as good as laying down 3. a simple presence detection model would solve that and also the algorithm already covers this