EMAN: Optimization-Driven Capacity Growth through Path Emergence in Multi-Task Learning
EMAN: Optimization-Driven Capacity Growth through Path Emergence in Multi-Task Learning
EMAN：多任务学习中通过路径涌现实现优化驱动的容量增长
Abstract: Existing multi-task learning methods rely on hard sharing, multiple paths or experts, adaptive sharing, and dynamic expansion. However, their capacity changes are usually constrained by predefined structures or triggered by task boundaries and conflict signals. This raises a fundamental question: can a network start from exact single-path computation and grow a new independent path only when persistent optimization evidence appears?
摘要： 现有的多任务学习方法主要依赖于硬共享、多路径或专家模型、自适应共享以及动态扩展。然而，这些方法的容量变化通常受到预定义结构的限制，或是由任务边界和冲突信号所触发。这提出了一个根本性的问题：网络能否从单一路径计算开始，仅在出现持续的优化证据时才生长出新的独立路径？
We propose the Emergent Modular Atomic Network (EMAN), an optimization-driven framework for exposing an antisymmetric growth direction through latent relative phases without instantiating a second path, and for monitoring multiple decision signals during training to transform local optimization evidence into a structural decision.
我们提出了涌现模块化原子网络（Emergent Modular Atomic Network, EMAN）。这是一个优化驱动的框架，它通过潜在的相对相位揭示反对称的增长方向，而无需实例化第二条路径；同时，它在训练过程中监控多个决策信号，将局部优化证据转化为结构性决策。
EMAN materializes two equal-capacity independent paths only after certification. EMAN adaptively allocates shared and task-specific representation capacity to accommodate varying task requirements. Extensive experiments on controlled rank settings, PASCAL-Context, and NYUv2 validate its effectiveness, achieving improved performance at a competitive computational cost.
EMAN 仅在经过验证后才会实现两条等容量的独立路径。EMAN 能够自适应地分配共享和任务特定的表示容量，以适应不同的任务需求。在受控秩设置、PASCAL-Context 和 NYUv2 数据集上的大量实验验证了其有效性，在具有竞争力的计算成本下实现了性能提升。
Paper Details:
- Authors: Chenlei Fang, Jingchen Li, Hongzong LI, Qingyao Li, Yixuan Zhang, Huarui Wu, Haobin Shi, Chunjiang Zhao
- arXiv ID: 2608.16930
- Subjects: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
论文详情：
- 作者： Chenlei Fang, Jingchen Li, Hongzong LI, Qingyao Li, Yixuan Zhang, Huarui Wu, Haobin Shi, Chunjiang Zhao
- arXiv ID: 2608.16930
- 学科分类： 机器学习 (cs.LG)；人工智能 (cs.AI)