Trust_Region_Methods.pdf
jubayer-ibn-hamid.github.io · 3,450 words · saved by 1 readers
N/A
Trust Region Methods Jubayer Ibn Hamid 1 Introduction These are notes on trust region optimization methods that I took as I read [1]. Intuitively, line search methods first make a quadratic model of the function f to generate a step direction and calculate the step length (preferably one that satisfies the Wolfe Conditions). In contrast, trust region methods also generate a model of the function f - but they define a region around where we are currently such that inside that region we believe our model is more or less the same as function f and then…
saved by
related reading
- Understanding Policy Gradients | John Lambertjohnwlambert.github.io
- Gregory Gundersengregorygundersen.com
- Why Momentum Really Worksdistill.pub
- [1605.08101] Global rates of convergence for nonconvex optimization on manifoldsarxiv.org
- prox_algs.pdfweb.stanford.edu
- Pen and Paper Exercises in Machine Learningarxiv.org
- Mathematical optimization - Wikipediaen.wikipedia.org
- bv_cvxbook.pdfweb.stanford.edu
- Universal Complexity Bounds for Universal Gradient Methods in Nonlinear Optimizationarxiv.org
- eecs127_reader.pdfeecs127.github.io
- On the Link Between Polynomials and Optimization, Part 1fa.bianp.net
- Gradient descent - Wikipediaen.wikipedia.org