Neural Parameter Allocation Search
This addresses memory and communication costs for neural network practitioners, offering a novel approach to parameter sharing that is more generalizable than hand-crafted methods.
The paper tackles the problem of high memory requirements in neural network training by introducing Neural Parameter Allocation Search (NPAS), a framework for training networks with fixed parameter budgets, and Shapeshifter Networks (SSNs) that automatically learn parameter sharing strategies, achieving performance gains without increasing inference FLOPs across tasks like ImageNet classification.
Training neural networks requires increasing amounts of memory. Parameter sharing can reduce memory and communication costs, but existing methods assume networks have many identical layers and utilize hand-crafted sharing strategies that fail to generalize. We introduce Neural Parameter Allocation Search (NPAS), a novel task where the goal is to train a neural network given an arbitrary, fixed parameter budget. NPAS covers both low-budget regimes, which produce compact networks, as well as a novel high-budget regime, where additional capacity can be added to boost performance without increasing inference FLOPs. To address NPAS, we introduce Shapeshifter Networks (SSNs), which automatically learn where and how to share parameters in a network to support any parameter budget without requiring any changes to the architecture or loss function. NPAS and SSNs provide a complete framework for addressing generalized parameter sharing, and can also be combined with prior work for additional performance gains. We demonstrate the effectiveness of our approach using nine network architectures across four diverse tasks, including ImageNet classification and transformers.