Deep Learning for Combinatorial Optimization - Count Your Flops & Make Your Flops Count
Institute for Pure & Applied Mathematics (IPAM) via YouTube
Overview
Explore a 23-minute lecture on applying deep learning to combinatorial optimization problems. Delve into the fundamental differences between combinatorial optimization and traditional machine learning tasks, and understand the trade-offs between computation reduction and solution quality. Learn about the importance of strategic model application, with practical examples illustrating how to balance the use of learned models and search algorithms. Gain insights into challenges and guidelines for future research directions in this field, presented by Wouter Kool from the University of Amsterdam. Cover topics including machine translation, neural network examples, dynamic programming, and the advantages and results of deep learning approaches in combinatorial optimization.
Syllabus
Introduction
Machine Translation vs Combinatorial Optimization
Neural Network Example
Dynamic Programming
Advantages
Results
Taught by
Institute for Pure & Applied Mathematics (IPAM)