I want to hack my car's climate control system, make it a little #smarter. Look, I use the #car to go places. And it usually includes getting out of the car and being outside for some while. So most of the time I dress according to the weather: in winter I put a warm coat and in summer I'm in no more than a T-shirt. And it can be not only seasonal. Some days are cooler or hotter than others. Even mornings can be much different that the afternoons. The thermostat that keeps the same temperature in the car's cabin no matter the weather outside is pretty dumb. It could be smarter than that. The car is already equipped with all necessary sensors and processing power.

Here is the idea. The thermostat doesn't have to keep the temperature precisely at the setpoint. When it's cool outside, which means I wear a sweater (likely), the temp can be somewhat below the setpoint. Conversely, in a hot summer day it's OK if the temperature is slightly higher.

Now, how would I program the thermostat? First of all it needs the reading from the outside thermometer. Which the car, I'm sure, has. This reading could add bias to the current setpoint that was set manually, almost following the outside temperature when it's not too far from the setpoint, but not too much off if it's really freezing or hot outside. One approach that I can think of is a #sigmoid function. The basic sigmoid, the logistic function needs 3 parameters, well, 2 if I fix the translation to 0, and my x will be the diff between the outside temperature and the #setpoint. Or even one parameter, the max allowed bias, if I fix the slope to 1 at x=0. This one parameter is some kind of "laziness" of the #thermostat that can be stored in settings.

๐Ÿค”

The Sigmoids Won't Save You

...

Astral Codex Ten

Hello World,๐Ÿ‘‹ #Introduction #NeuHier

I'm Esther, working in insurance. No, it's not as boring as it sounds. I just moved here from the "other place" to have more in-depth conversations about AI Ethics, Data Privacy and the joy and pain of training Small LLMs without a huge tech stack.

Excited to learnfrom the #Sigmoid Community and connect with humans, not bots.

Hallo Fediverse! Ich freue mich auf den Austausch zu KI-Ethik, Governance und allem, was Tech besser macht.

I really like darktable 5.4. Wayland support works very nicely. Small improvements like higher contrast module toggles to know what's turned on or off, is helpful.

Great work devs!

I think (for now) I'll continue to use Sigmoid and Filmic RGB tone mappers.

AgX shows real promise, and I'll certainly give it another try with the next update. Just can't get any satisfaction with the skies.

#darktable #AgX #Sigmoid #FilmicRGB

Iโ€™m MetaDarko, just moved here from mastodon.uno.
I share music, stories, and the little projects Iโ€™m working on.

Thanks for the welcome Iโ€™m looking forward to chatting with everyone! ๐Ÿ˜€

#sigmoid #music #projects #community #hi #art #ai #friends

Logistic regression may be used for classification.

In order to preserve the convex nature for the loss function, a log-loss cost function has been designed for logistic regression. This cost function extremes at labels True and False.

The gradient for the loss function of logistic regression comes out to have the same form of terms as the gradient for the Least Squared Error.

More: https://www.baeldung.com/cs/gradient-descent-logistic-regression

#optimization #algebra #linearAlgebra #math #maths #mathematics #mathStodon #ML #dataScience #machineLearning #DeepLearning #neuralNetworks #NLP #modeling #modelling #models #dataDev #AIDev #regression #modelling #dataLearning #probabilities #logisticRegression #logLoss #sigmoid #classification #differentialCalculus #loss

I'm still experimenting with #sigmoid in #darktable to achieve a more #AnaloguePhotography look and feel. This was taken in September 2020 at #ExtinctionRebellion in #London I think it works quite well #photo #photography
I'm sure lots of #photographers out there use #darktable what do people think? Can you get the results you want? Do you have a favourite workflow? I used #sigmoid to process this #photo but i normally use #filmic I also used #ColorLookUpTable to apply #FujiVelviaEmulation It gives it a very #analogue feel, but sometimes when you look at things too long it becomes really hard to tell what is too much. I think I quite like it, but maybe tomorrow I'll be less sure.

@ai6yr
Failure to have a $(HotTopic) strategy is bad for shareholder value, so having a strategy must be good.

Data harvesting/resale. Blockchain. CraptoCoyn. NFT. LLM/^AI^/ ChatBot.

Being a waster of money, electricity, fossil carbon is apparently irrelevant to 90-day long term planning of quarterly share holder value forever growth.
(#sigmoid curves look like exponential until they don't. )