#ML note2self:From ref 1> we can take any expected utility maximization problem, and decompose it into an entropy minimization term plus a “make-the-world-look-like-this-specific-model” term.This view should be combined with ref 2. If the utility is related to the curvature of the discrete state space, we are making a connection between entropy + KL divergence and curvature on graph. (This idea has to be polished in depth.)Refs:1. Trivial proof but interesting perspective:

#ML

note2self:

From ref 1

> we can take any expected utility maximization problem, and decompose it into an entropy minimization term plus a “make-the-world-look-like-this-specific-model” term.

This view should be combined with ref 2. If the utility is related to the curvature of the discrete state space, we are making a connection between entropy + KL divergence and curvature on graph. (This idea has to be polished in depth.)

Refs:
1. Trivial proof but interesting perspective: https://www.lesswrong.com/posts/voLHQgNncnjjgAPH7/utility-maximization-description-length-minimization
2. Samal Areejit, Pharasi Hirdesh K., Ramaia Sarath Jyotsna, Kannan Harish, Saucan Emil, Jost Jürgen and Chakraborti Anirban 2021Network geometry and market instabilityR. Soc. open sci.8201734. http://doi.org/10.1098/rsos.201734

Lesswrong

Utility Maximization = Description Length Minimization — LessWrong

There’s a useful intuitive notion of “optimization” as pushing the world into a small set of states, starting from any of a large number of states. V…