Fishers for Free? Approximating the Fisher Information Matrix by Recycling the Squared Gradient Accumulator

• URL: http://arxiv.org/abs/2507.18807v1
• Date: Thu, 24 Jul 2025 21:10:37 GMT
• Title: Fishers for Free? Approximating the Fisher Information Matrix by Recycling the Squared Gradient Accumulator
• Authors: YuXin Li, Felix Dangel, Derek Tam, Colin Raffel,
• Abstract summary: The diagonal of a model's Fisher Information Matrix (the "Fisher diagonal") has frequently been used as a way to measure parameter sensitivity.<n>This paper explores whether an approximation of the Fisher diagonal can be obtained "for free" by recycling the squared gradient that has already been computed over the course of training.
• Score: 39.369412319701006
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: The diagonal of a model's Fisher Information Matrix (the "Fisher diagonal") has frequently been used as a way to measure parameter sensitivity. Typically, the Fisher diagonal is estimated via squared sampled gradients of the model's likelihood with respect to its parameters, averaged over a few hundred or thousand examples -- a process which incurs nontrivial computational costs. At the same time, adaptive gradient methods like the ubiquitous Adam optimizer compute a moving average of the squared gradient over the course of training. This paper therefore explores whether an approximation of the Fisher diagonal can be obtained "for free" by recycling the squared gradient accumulator that has already been computed over the course of training. Through a comprehensive set of experiments covering five applications of the Fisher diagonal, we demonstrate that the "Squisher" (SQUared gradient accumulator as an approximation of the FISHER) consistently performs similarly to the Fisher diagonal while outperforming baseline methods. Additionally, we clarify the exact differences between the Squisher and the Fisher diagonal and provide empirical quantification of their respective impact.

Related papers

• Generalized Fisher-Weighted SVD: Scalable Kronecker-Factored Fisher Approximation for Compressing Large Language Models [6.57101653042078]
  Generalized Fisher-Weighted SVD (GFWSVD) is a post-training compression technique that accounts for both diagonal and off-diagonal elements of the Fisher information matrix.<n>We demonstrate the effectiveness of our method on LLM compression, showing improvements over existing compression baselines.
  arXiv  Detail & Related papers  (2025-05-23T14:41:52Z)
• Approximation and bounding techniques for the Fisher-Rao distances between parametric statistical models [7.070726553564701]
  We consider several numerically robust approximation and bounding techniques for the Fisher-Rao distances.<n>In particular, we obtain a generic method to guarantee an arbitrarily small additive error on the approximation.<n>We propose two new distances based either on the Fisher-Rao lengths of curves serving as proxies of Fisher-Rao geodesics.
  arXiv  Detail & Related papers  (2024-03-15T08:05:16Z)
• Closed-form Filtering for Non-linear Systems [83.91296397912218]
  We propose a new class of filters based on Gaussian PSD Models, which offer several advantages in terms of density approximation and computational efficiency. We show that filtering can be efficiently performed in closed form when transitions and observations are Gaussian PSD Models. Our proposed estimator enjoys strong theoretical guarantees, with estimation error that depends on the quality of the approximation and is adaptive to the regularity of the transition probabilities.
  arXiv  Detail & Related papers  (2024-02-15T08:51:49Z)
• Unbiased Kinetic Langevin Monte Carlo with Inexact Gradients [0.8749675983608172]
  We present an unbiased method for posterior means based on kinetic Langevin dynamics. Our proposed estimator is unbiased, attains finite variance, and satisfies a central limit theorem. Our results demonstrate that in large-scale applications, the unbiased algorithm we present can be 2-3 orders of magnitude more efficient than the gold-standard" randomized Hamiltonian Monte Carlo.
  arXiv  Detail & Related papers  (2023-11-08T21:19:52Z)
• Low-rank extended Kalman filtering for online learning of neural networks from streaming data [71.97861600347959]
  We propose an efficient online approximate Bayesian inference algorithm for estimating the parameters of a nonlinear function from a potentially non-stationary data stream. The method is based on the extended Kalman filter (EKF), but uses a novel low-rank plus diagonal decomposition of the posterior matrix. In contrast to methods based on variational inference, our method is fully deterministic, and does not require step-size tuning.
  arXiv  Detail & Related papers  (2023-05-31T03:48:49Z)
• Natural Gradient Methods: Perspectives, Efficient-Scalable Approximations, and Analysis [0.0]
  Natural Gradient Descent is a second-degree optimization method motivated by the information geometry. It makes use of the Fisher Information Matrix instead of the Hessian which is typically used. Being a second-order method makes it infeasible to be used directly in problems with a huge number of parameters and data.
  arXiv  Detail & Related papers  (2023-03-06T04:03:56Z)
• Memory-Efficient Backpropagation through Large Linear Layers [107.20037639738433]
  In modern neural networks like Transformers, linear layers require significant memory to store activations during backward pass. This study proposes a memory reduction approach to perform backpropagation through linear layers.
  arXiv  Detail & Related papers  (2022-01-31T13:02:41Z)
• Learning Linearized Assignment Flows for Image Labeling [70.540936204654]
  We introduce a novel algorithm for estimating optimal parameters of linearized assignment flows for image labeling. We show how to efficiently evaluate this formula using a Krylov subspace and a low-rank approximation.
  arXiv  Detail & Related papers  (2021-08-02T13:38:09Z)
• Two-Level K-FAC Preconditioning for Deep Learning [7.699428789159717]
  In the context of deep learning, many optimization methods use gradient covariance information in order to accelerate the convergence of Gradient Descent. In particular, starting with Adagrad, a seemingly endless line of research advocates the use of diagonal approximations of the so-called empirical Fisher matrix. One particularly successful variant of such methods is the so-called K-FAC, which uses a Kronecker-ed block-factored preconditioner.
  arXiv  Detail & Related papers  (2020-11-01T17:54:21Z)
• Path Sample-Analytic Gradient Estimators for Stochastic Binary Networks [78.76880041670904]
  In neural networks with binary activations and or binary weights the training by gradient descent is complicated. We propose a new method for this estimation problem combining sampling and analytic approximation steps. We experimentally show higher accuracy in gradient estimation and demonstrate a more stable and better performing training in deep convolutional models.
  arXiv  Detail & Related papers  (2020-06-04T21:51:21Z)

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.