Caching for Edge Inference at Scale: A Mean Field Multi-Agent Reinforcement Learning Approach | IEEE Conference Publication | IEEE Xplore