A Neural Scaling Law from Lottery Ticket Ensembling

• URL: http://arxiv.org/abs/2310.02258v2
• Date: Fri, 2 Feb 2024 02:09:51 GMT
• Title: A Neural Scaling Law from Lottery Ticket Ensembling
• Authors: Ziming Liu, Max Tegmark
• Abstract summary: Sharma & Kaplan predicted that MSE losses decay as $N-alpha$, $alpha=4/d$, where $N$ is the number of model parameters, and $d$ is the intrinsic input dimension. We find that a simple 1D problem manifests a different scaling law ($alpha=1$) from their predictions.
• Score: 19.937894875216507
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Neural scaling laws (NSL) refer to the phenomenon where model performance improves with scale. Sharma & Kaplan analyzed NSL using approximation theory and predict that MSE losses decay as $N^{-\alpha}$, $\alpha=4/d$, where $N$ is the number of model parameters, and $d$ is the intrinsic input dimension. Although their theory works well for some cases (e.g., ReLU networks), we surprisingly find that a simple 1D problem $y=x^2$ manifests a different scaling law ($\alpha=1$) from their predictions ($\alpha=4$). We opened the neural networks and found that the new scaling law originates from lottery ticket ensembling: a wider network on average has more "lottery tickets", which are ensembled to reduce the variance of outputs. We support the ensembling mechanism by mechanistically interpreting single neural networks, as well as studying them statistically. We attribute the $N^{-1}$ scaling law to the "central limit theorem" of lottery tickets. Finally, we discuss its potential implications for large language models and statistical physics-type theories of learning.

Related papers

• Scaling Laws in Linear Regression: Compute, Parameters, and Data [86.48154162485712]
  We study the theory of scaling laws in an infinite dimensional linear regression setup. We show that the reducible part of the test error is $Theta(-(a-1) + N-(a-1)/a)$. Our theory is consistent with the empirical neural scaling laws and verified by numerical simulation.
  arXiv  Detail & Related papers  (2024-06-12T17:53:29Z)
• Language models scale reliably with over-training and on downstream tasks [121.69867718185125]
  Scaling laws are useful guides for derisking expensive training runs. However, there remain gaps between current studies and how language models are trained. In contrast, scaling laws mostly predict loss on inference, but models are usually compared on downstream task performance.
  arXiv  Detail & Related papers  (2024-03-13T13:54:00Z)
• A Dynamical Model of Neural Scaling Laws [79.59705237659547]
  We analyze a random feature model trained with gradient descent as a solvable model of network training and generalization. Our theory shows how the gap between training and test loss can gradually build up over time due to repeated reuse of data.
  arXiv  Detail & Related papers  (2024-02-02T01:41:38Z)
• Generalization and Stability of Interpolating Neural Networks with Minimal Width [37.908159361149835]
  We investigate the generalization and optimization of shallow neural-networks trained by gradient in the interpolating regime. We prove the training loss number minimizations $m=Omega(log4 (n))$ neurons and neurons $Tapprox n$. With $m=Omega(log4 (n))$ neurons and $Tapprox n$, we bound the test loss training by $tildeO (1/)$.
  arXiv  Detail & Related papers  (2023-02-18T05:06:15Z)
• The Onset of Variance-Limited Behavior for Networks in the Lazy and Rich Regimes [75.59720049837459]
  We study the transition from infinite-width behavior to this variance limited regime as a function of sample size $P$ and network width $N$. We find that finite-size effects can become relevant for very small datasets on the order of $P* sim sqrtN$ for regression with ReLU networks.
  arXiv  Detail & Related papers  (2022-12-23T04:48:04Z)
• Minimax Optimal Quantization of Linear Models: Information-Theoretic Limits and Efficient Algorithms [59.724977092582535]
  We consider the problem of quantizing a linear model learned from measurements. We derive an information-theoretic lower bound for the minimax risk under this setting. We show that our method and upper-bounds can be extended for two-layer ReLU neural networks.
  arXiv  Detail & Related papers  (2022-02-23T02:39:04Z)
• On the Provable Generalization of Recurrent Neural Networks [7.115768009778412]
  We analyze the training and generalization for Recurrent Neural Network (RNN) We prove a generalization error bound to learn functions without normalized conditions. We also prove a novel result to learn N-variables functions of input sequence.
  arXiv  Detail & Related papers  (2021-09-29T02:06:33Z)
• Locality defeats the curse of dimensionality in convolutional teacher-student scenarios [69.2027612631023]
  We show that locality is key in determining the learning curve exponent $beta$. We conclude by proving, using a natural assumption, that performing kernel regression with a ridge that decreases with the size of the training set leads to similar learning curve exponents to those we obtain in the ridgeless case.
  arXiv  Detail & Related papers  (2021-06-16T08:27:31Z)
• Learning Curve Theory [21.574781022415365]
  Scaling laws' refers to power-law decreases of training or test error w.r.t. more data, larger neural networks, and/or more compute. We develop and theoretically analyse the simplest possible (toy) model that can exhibit $n-beta$ learning curves for arbitrary power $beta>0$.
  arXiv  Detail & Related papers  (2021-02-08T09:25:31Z)
• A Neural Scaling Law from the Dimension of the Data Manifold [8.656787568717252]
  When data is plentiful, the loss achieved by well-trained neural networks scales as a power-law $L propto N-alpha$ in the number of network parameters $N$. The scaling law can be explained if neural models are effectively just performing regression on a data manifold of intrinsic dimension $d$. This simple theory predicts that the scaling exponents $alpha approx 4/d$ for cross-entropy and mean-squared error losses.
  arXiv  Detail & Related papers  (2020-04-22T19:16:06Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.