Deep learning-assisted heuristic tree search (DLTS) combines a heuristic tree-search algorithm with deep neural networks (DNNs). The networks repeatedly inspect the algorithm’s current state and guide which branches to explore and how to bound nodes. section 11.5
Its defining workflow includes:
- Representing the current algorithm state, potentially including the problem definition, as model input. section 11.5
- Reusing the same ML model across repeated combinatorial-optimization decisions. section 11.5
- Using DNN predictions to select branches and determine node bounds. section 11.5
Figure 11.14 Machine learning (ML) for combinatorial optimization (CO) problems
Maps machine learning methods to combinatorial-optimization problems and situates DLTS within that broader relationship. section 11.5
