[ 3 / biz / cgl / ck / diy / fa / ic / jp / lit / sci / vr / vt ] [ index / top / reports ] [ become a patron ] [ status ]
2023-11: Warosu is now out of extended maintenance.

/sci/ - Science & Math


View post   

File: 6 KB, 275x183, frist.jpg [View same] [iqdb] [saucenao] [google]
9143701 No.9143701 [Reply] [Original]

Can anyone explain to me what energy is in machine learning and the idea of minima and how it relates to learning, models etc.

More so, i'm also wondering about the whole relationship betwen thermodynamics and machine learning. Whats that about?

>> No.9143784

>>9143701
The only parallel (I'm aware of) between ML and thermodynamics is the concept of entropy.

Entropy, being (informally) a measure of the number of (or more precisely, distribution of) feasible "states" or "configurations" of a system translates naturally to a form of uncertainty quantification in ML.

You may wish to have a look at "Shannon Entropy" if this interests you.

>> No.9144071

>>9143784
Other parallels are the boltzmann distribution. And all neural networks seem to be quantifiable by an energy which seems to have been borrowed from chemical physics (?).