LGDIS-NNMLFeb 12, 2021

Explaining Neural Scaling Laws

arXiv:2102.06701v2457 citations
AI Analysis

This work addresses the fundamental understanding of scaling laws in deep learning for researchers, offering a taxonomy and insights into mechanisms driving loss improvements, though it is incremental in building on existing observations.

The authors tackled the problem of explaining the power-law scaling relations observed in neural network population loss with dataset or model size, identifying four distinct scaling regimes and providing a theoretical framework that connects them, supported by empirical validation on standard architectures and datasets.

The population loss of trained deep neural networks often follows precise power-law scaling relations with either the size of the training dataset or the number of parameters in the network. We propose a theory that explains the origins of and connects these scaling laws. We identify variance-limited and resolution-limited scaling behavior for both dataset and model size, for a total of four scaling regimes. The variance-limited scaling follows simply from the existence of a well-behaved infinite data or infinite width limit, while the resolution-limited regime can be explained by positing that models are effectively resolving a smooth data manifold. In the large width limit, this can be equivalently obtained from the spectrum of certain kernels, and we present evidence that large width and large dataset resolution-limited scaling exponents are related by a duality. We exhibit all four scaling regimes in the controlled setting of large random feature and pretrained models and test the predictions empirically on a range of standard architectures and datasets. We also observe several empirical relationships between datasets and scaling exponents under modifications of task and architecture aspect ratio. Our work provides a taxonomy for classifying different scaling regimes, underscores that there can be different mechanisms driving improvements in loss, and lends insight into the microscopic origins of and relationships between scaling exponents.

Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes