Neural Parameter Allocation Search

TL;DR

This paper introduces NPAS, a framework for training neural networks within arbitrary parameter budgets, and proposes SSNs that automatically learn parameter sharing to optimize network capacity and performance.

Contribution

The paper presents NPAS and SSNs, enabling flexible parameter sharing in neural networks without architecture changes, covering both low and high-budget regimes.

Findings

Effective across nine architectures and four tasks.

Supports both compact and high-capacity networks.

Improves performance without increasing inference FLOPs.

Abstract

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.