Bounded Curation: Optimization — Research — II
This post is a long due follow up on the first of the research posts of the bounded curation series.
Boyd & Vanderberghe’s Lecture
Boyd’s lecture on convex optimization based on Boyd and Vanderberghe’s book. Material includes the book, lecture slides (summer 2023 update) and exercises.
Generative deep learning for decision making in gas networks
This article uses GANs (Generative Adversarial Networks) as a decision support system to MILP formulations. They feed problem instances (same model different input parameters) to the GAN as training data to reduce problem solving times by a significant factor.
Augmenting High-dimensional Nonlinear Optimization with Conditional GANs
Another GAN paper on optimization. This time a Genetic Algorithm (GA) is used to generate solutions to a high-dimensional nonlinear problem. The paper has a Github Repository
OR StackExchange: How much can we expect to increase the speed of mixed integer programming in the next 10 years?
Interesting discussion on the future of MILP. It is about speed improvements, performance benchmarking and machine learning. Some mixed forecasts but definitely worth a read.
PhD Thesis (Yunhao Tang) - Reinforcement Learning: New Algorithms and An Application for Integer Programming
In my opinion Reinforcement Learning methods can be used in/as solvers. Though it is not as efficient yet. This PhD thesis explores RL’s abilities to learn an efficient cutting algorithm.
Neurips Paper (Dai et al.) — Learning Combinatorial Optimization Algorithms over Graphs
Learning CO problems with a mixture of RL and graph methods to generate a greedy heuristic.
Still a few papers to go… See you on part III.
I personally use these posts to wrap up my research on specific topics and I make them available to public for the benefit of the wider audience.
