NISHIO Hirokazu[English][日本語]

Introduction to Information-Geometric Optimization

Information-Geometric Optimization Cybozu Labs Study Session image - GA - evolutionary strategy

image image - optimization problem

  • No requirement that the objective function be differentiable
    • NN is a bundle of differentiable functions
  • Interpret genetic algorithms as updating probability distributions

image

  • The general gradient method differentiates the objective function f on X-space and updates it using the gradient
  • Objective function is not differentiable
  • Let's make a differentiable objective function g on the parameter interval

image

  • Means of creating functions on parameter space
  • Example: Average
  • IGO "I don't like this."

image image image image image ---- before any additions were made during the study session.

image - GA - evolutionary strategy image image - optimization problem

  • No requirement that the objective function be differentiable
    • NN is a bundle of differentiable functions
  • Interpret genetic algorithms as updating probability distributions image
  • The general gradient method differentiates the objective function f on X-space and updates it using the gradient
  • Objective function is not differentiable
  • Let's make a differentiable objective function g on the parameter interval image
  • Means of creating functions on parameter space
  • Example: Average
  • IGO "I don't like this." image image image image image

--- Preparation of documents image image image image image image image image image image image image


This page is auto-translated from [/nishio/Information-Geometric Optimizationの紹介](https://scrapbox.io/nishio/Information-Geometric Optimizationの紹介) using DeepL. If you looks something interesting but the auto-translated English is not good enough to understand it, feel free to let me know at @nishio_en. I'm very happy to spread my thought to non-Japanese readers.


(C)NISHIO Hirokazu / Converted from Markdown (en)
Source: [GitHub] / [Scrapbox]