# Optimizing Memory Placement using Evolutionary Graph Reinforcement Learning

arXiv: https://arxiv.org/abs/2007.07298 · HTML: https://mlsanity.com/paper/2007.07298/

## Key facts

- Cited by 7 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 14 Jul 2020.

## Built on by (most cited first, 7 of 7)

- [Robust Scheduling with GFlowNets](https://mlsanity.com/paper/2302.05446/) (2023)
- [A Survey of Machine Learning for Computer Architecture and Systems](https://mlsanity.com/paper/2102.07952/) (2021)
- [Searching for High-Value Molecules Using Reinforcement Learning and Transformers](https://mlsanity.com/paper/2310.02902/) (2023)
- [Neuroevolution-Enhanced Multi-Objective Optimization for Mixed-Precision Quantization](https://mlsanity.com/paper/2106.07611/) (2021)
- [RL4ReAl: Reinforcement Learning for Register Allocation](https://mlsanity.com/paper/2204.02013/) (2022)
- [Partitioning Distributed Compute Jobs with Reinforcement Learning and Graph Neural Networks](https://mlsanity.com/paper/2301.13799/) (2023)
- [Diverse LLMs or Diverse Question Interpretations? That is the Ensembling Question](https://mlsanity.com/paper/2507.21168/) (2025)

## Abstract (arXiv)

For deep neural network accelerators, memory movement is both energetically expensive and can bound computation. Therefore, optimal mapping of tensors to memory hierarchies is critical to performance. The growing complexity of neural networks calls for automated memory mapping instead of manual heuristic approaches; yet the search space of neural network computational graphs have previously been prohibitively large. We introduce Evolutionary Graph Reinforcement Learning (EGRL), a method designed for large search spaces, that combines graph neural networks, reinforcement learning, and evolutionary search. A set of fast, stateless policies guide the evolutionary search to improve its sample-efficiency. We train and validate our approach directly on the Intel NNP-I chip for inference. EGRL outperforms policy-gradient, evolutionary search and dynamic programming baselines on BERT, ResNet-101 and ResNet-50. We additionally achieve 28-78% speed-up compared to the native NNP-I compiler on all three workloads.