such is life

"we know the risk of the outcome is very low" sounds like prior information to me - are Bayesian approaches on the table?

If not, I imagine some penalty could give similar results? But in either case, if getting more data is a reasonable ask, that'd likely be better regardless?

bleak house does, in fact, have the juice, i fear

The TLDR is that I set up a process on a server I have (that has a Bluetooth adapter) to listen for my controller scanning for a connection and waking the gaming PC via Wake-on-LAN. Pretty happy with it.

affiliate marketing is the will of god

The results for my ThinkPad are that each percent increase in screen brightness increases power by about 22 mW. Some ballpark calculations show a decrease in total battery life by about 10 minutes for every 4% increase in brightness.

Fun little experiment to quantify something I've wondered about!

I fit a very simple Bayesian model in Stan to infer the direct effect of screen brightness on power draw

👋

Computing avg power draw via change in energy / change in time still works! But importantly, on my machine at least, the energy readings have much lower fidelity over short time scales than power!

It does mean that if one were to try to perform any experimentation that used power observations from the battery as estimates of real-time power draw, it would be thrown off unless the experiments were run over long time scale.

I'm sure this behavior is well-known to some, but it was new to me (and I haven't been able to find any references to this smoothing process, I assume its hardware specific?).

In retrospect, it's sensible that smoothing is applied given the common use case of estimating remaining battery time.

Thanks to the simple form of the exponential smoothing applied here, I can invert the smoothing to recover the latent true power readings and get a much more realistic looking profile:

In the post, I show that the observed readings are highly consistent with an exponential smoothing process being applied to latent "true" readings. By assuming that the latent power profile is a step function, I find a smoothing parameter that maps the idealized latent power to the observations.

By taking readings from the battery while applying a step-function artificial load, I observed that the reported power decayed toward near-constant values in a manner that appeared exponential

A red kite, perhaps?

being a longtime python user, my perception of julia is that I can see myself really liking it, and I appreciate many of it's ideas, but python has never really given me a strong enough reason to switch. especially when it feels like python is making progress at improving itself

ah yes, the 'decent into madness' phase of becoming a Bayesian

I think I've been to that show. kicked ass

me: "ok but what if the reader isn't very good"