MaRCA: Multi‑Agent Reinforcement Learning Computation Allocation for Full‑Chain Ad Serving

This article presents MaRCA, a multi‑agent reinforcement learning framework that allocates computation resources across the full ad‑serving chain by modeling user value, compute consumption, and action rewards, enabling fine‑grained power‑tilting toward high‑quality traffic and achieving significant business gains under strict latency constraints.

AI Optimizationad servingcomputation allocation

0 likes · 16 min read

MaRCA: Multi‑Agent Reinforcement Learning Computation Allocation for Full‑Chain Ad Serving

21CTO

Jul 1, 2021 · Operations

How Meituan Optimized Delivery Ad Compute with Intelligent Power Allocation

This article describes Meituan's delivery advertising system's shift to intelligent compute allocation, detailing the business challenges, the four‑element framework for dynamic resource distribution, the optimal tier decision process, system stability mechanisms, experimental results, and future research directions.

MeituanOnline Optimizationad serving

0 likes · 19 min read

How Meituan Optimized Delivery Ad Compute with Intelligent Power Allocation

Meituan Technology Team

Jun 17, 2021 · Artificial Intelligence

Intelligent Compute Allocation for Meituan Takeaway Advertising: Design, Implementation, and Experiments

The paper introduces an intelligent compute framework for Meituan Takeaway’s ad platform that dynamically allocates CPU/GPU resources to traffic based on quantified business value, using elastic queues, models, channels and links, optimizing tiers via offline λ search and online PID‑controlled decisions, achieving a 2.3 % revenue lift with unchanged resources and a 40 % resource cut while preserving revenue.

ad servingintelligent computeoptimization

0 likes · 18 min read

Intelligent Compute Allocation for Meituan Takeaway Advertising: Design, Implementation, and Experiments