I see there was a thread about this topic in 2017, but further commenting is closed. Why close a thread if this is a recurring issue that has not yet been fixed. I would think Sense would like its user community to report such issues. I did go to my Oven device in the Sense app and informed Sense this device was not on. Shame there was no provision for me to inform Sense what device (Dryer) was actually one. I don’t buy the Sense argument as posted by BenATSense that Sense has trouble telling the difference between “heating devices”. A typical oven has no motor, whereas a dryer has a motor, so these appliances should be clearly differentiated by their signals. Agree that some may have a convection oven, but that is a function we do not use, so Sense should not be confusing these devices.
I think you misunderstand BenatSense’s comment, or might be missing a subtlety here. All Sense sees is waveforms, and complex devices like washers, drivers, even fridges have composite waveforms consisting of many different components. There’s nothing in the house waveform that links the running of the dryer motor/tumbler with the start and stop of the heating elements. A dryer heating element waveform looks a lot like an oven heating element waveform.
Keep in mind the system on Discourse May have automatically closed the thread. It does close threads after a year of not having a comment. I don’t think sense is trying to close comments to shit down the conversation. They are very well aware this still happens.
Kevin1 - Agreed there is nothing in the house wave form that intrinsically links the starting of a motor followed by a heating element, but that is in fact the whole point of the AI is to spot and make such linkages through strengthening of the neural net associations. Hence it is rather surprising to me that Sense’s AI has not long ago figured out how to discern an dryer composite waveform from an oven operating to a near square-wave without any preceding motor start spike. One should never let the AI operate without oversight and tuning, else you end up with the Buddhist monk problem (everything is connected) or worse the Microsoft Tay Bot. If Sense does have such confusion, then the app should enable a user to quickly and easily report the error and inform Sense of correct appliance.
Agree on the need to improve reporting for conflated devices. But I do think your expectations for AI are a bit optimistic for electrical disaggregation.
Threads without a response get closed after a year. We do that to keep this forum organized. Device conflation still happens and you’ll find many threads here that mention it. For issues of conflation, you should write into the Support team. Sometimes, they’re able to help force some differentiation in the backend.
Keep in mind that while it may be (relatively) simple to discern a dryer from an oven in a vacuum, your house has many other competing signals. Those competing signals create the biggest challenge for us. If you only ever ran one device at a time, discerning the oven from the dryer would be a much simpler problem to solve. Unfortunately for us, homes just aren’t that simple. We don’t natively see isolated signals, but messy composites of many signals that always look different. It’s like a needle in a haystack, but the haystack and needle keep changing in size and shape.
I’d recommend reading this thread, as it goes into some of the complexities of ML-based device disaggregation: Why can't you train Sense?
This topic was automatically closed 365 days after the last reply. New replies are no longer allowed.