English
Related papers

Related papers: Flatness is a False Friend

200 papers

Over-parameterized neural networks generalize well in practice without any explicit regularization. Although it has not been proven yet, empirical evidence suggests that implicit regularization plays a crucial role in deep learning and…

Machine Learning · Computer Science 2019-03-07 Masayoshi Kubo , Ryotaro Banno , Hidetaka Manabe , Masataka Minoji

Zeroth-order methods are extensively used in machine learning applications where gradients are infeasible or expensive to compute, such as black-box attacks, reinforcement learning, and language model fine-tuning. Existing optimization…

Machine Learning · Computer Science 2025-11-12 Liang Zhang , Bingcong Li , Kiran Koshy Thekumparampil , Sewoong Oh , Michael Muehlebach , Niao He

Recently, it has been observed that when training a deep neural net with SGD, the majority of the loss landscape's curvature quickly concentrates in a tiny *top* eigenspace of the loss Hessian, which remains largely stable thereafter.…

Machine Learning · Computer Science 2025-04-22 Andres Fernandez , Frank Schneider , Maren Mahsereci , Philipp Hennig

In recent years, various notions of capacity and complexity have been proposed for characterizing the generalization properties of stochastic gradient descent (SGD) in deep learning. Some of the popular notions that correlate well with the…

Optimization and Control · Mathematics 2021-06-15 Mert Gurbuzbalaban , Umut Şimşekli , Lingjiong Zhu

Federated Learning (FL) is an emerging machine learning framework that enables multiple clients (coordinated by a server) to collaboratively train a global model by aggregating the locally trained models without sharing any client's…

Machine Learning · Computer Science 2025-12-10 M Yashwanth , Gaurav Kumar Nayak , Harsh Rangwani , Arya Singh , R. Venkatesh Babu , Anirban Chakraborty

Weight normalization (WeightNorm) is widely used in practice for the training of deep neural networks and modern deep learning libraries have built-in implementations of it. In this paper, we provide the first theoretical characterizations…

Machine Learning · Computer Science 2025-01-22 Pedro Cisneros-Velarde , Zhijie Chen , Sanmi Koyejo , Arindam Banerjee

Meta-learning that uses implicit gradient have provided an exciting alternative to standard techniques which depend on the trajectory of the inner loop training. Implicit meta-learning (IML), however, require computing $2^{nd}$ order…

Machine Learning · Computer Science 2023-10-31 Fady Rezk

Despite the fact that the loss functions of deep neural networks are highly non-convex, gradient-based optimization algorithms converge to approximately the same performance from many random initial points. One thread of work has focused on…

Machine Learning · Computer Science 2020-03-24 Charles G. Frye , James Simon , Neha S. Wadia , Andrew Ligeralde , Michael R. DeWeese , Kristofer E. Bouchard

The Hessian of a neural network captures parameter interactions through second-order derivatives of the loss. It is a fundamental object of study, closely tied to various problems in deep learning, including model design, optimization, and…

Machine Learning · Computer Science 2021-07-02 Sidak Pal Singh , Gregor Bachmann , Thomas Hofmann

Weight decay is ubiquitous in training deep neural network architectures. Its empirical success is often attributed to capacity control; nonetheless, our theoretical understanding of its effect on the loss landscape and the set of…

Machine Learning · Statistics 2026-03-31 Anil Kamber , Rahul Parhi

Understanding the curvature evolution of the loss landscape is fundamental to analyzing the training dynamics of neural networks. The most commonly studied measure, Hessian sharpness ($\lambda_{\max}^H$) -- the largest eigenvalue of the…

Recent work has shown that methods like SAM which either explicitly or implicitly penalize second order information can improve generalization in deep learning. Seemingly similar methods like weight noise and gradient penalties often fail…

Machine Learning · Computer Science 2024-01-26 Yann N. Dauphin , Atish Agarwala , Hossein Mobahi

Normalization layers (e.g., Batch Normalization, Layer Normalization) were introduced to help with optimization difficulties in very deep nets, but they clearly also help generalization, even in not-so-deep nets. Motivated by the long-held…

Machine Learning · Computer Science 2023-01-18 Kaifeng Lyu , Zhiyuan Li , Sanjeev Arora

It is arguably believed that flatter minima can generalize better. However, it has been pointed out that the usual definitions of sharpness, which consider either the maxima or the integral of loss over a $\delta$ ball of parameters around…

Machine Learning · Computer Science 2021-01-11 Mingyang Yi , Huishuai Zhang , Wei Chen , Zhi-Ming Ma , Tie-Yan Liu

It was empirically confirmed by Keskar et al.\cite{SharpMinima} that flatter minima generalize better. However, for the popular ReLU network, sharp minimum can also generalize well \cite{SharpMinimacan}. The conclusion demonstrates that the…

Machine Learning · Computer Science 2019-03-07 Mingyang Yi , Qi Meng , Wei Chen , Zhi-ming Ma , Tie-Yan Liu

Neural networks are usually not the tool of choice for nonparametric high-dimensional problems where the number of input features is much larger than the number of observations. Though neural networks can approximate complex multivariate…

Methodology · Statistics 2019-06-25 Jean Feng , Noah Simon

The loss and the norm of its gradient separate the healthy and the pathological regimes of neural-network training only weakly, whilst the curvature of the empirical risk differs qualitatively between them but is inaccessible explicitly at…

Machine Learning · Computer Science 2026-05-26 Maxim Bolshim , Alexander Kugaevskikh

Near an optimal learning point of a neural network, the learning performance of gradient descent dynamics is dictated by the Hessian matrix of the loss function with respect to the network parameters. We characterize the Hessian…

Machine Learning · Statistics 2025-12-18 Carlos Couto , José Mourão , Mário A. T. Figueiredo , Pedro Ribeiro

In this work we study generalization of neural networks in gradient-based meta-learning by analyzing various properties of the objective landscapes. We experimentally demonstrate that as meta-training progresses, the meta-test solutions,…

Machine Learning · Computer Science 2019-07-18 Simon Guiroy , Vikas Verma , Christopher Pal

Training modern neural networks often relies on large learning rates, operating at the edge of stability, where the optimization dynamics exhibit oscillatory and chaotic behavior. Empirically, this regime often yields improved…

Machine Learning · Computer Science 2026-04-22 Mario Tuci , Caner Korkmaz , Umut Şimşekli , Tolga Birdal