中文
相关论文

相关论文: A Proof of Learning Rate Transfer under $\mu$P

200 篇论文

Transferring the optimal learning rate from small to large neural networks can enable efficient training at scales where hyperparameter tuning is otherwise prohibitively expensive. To this end, the Maximal Update Parameterization (muP)…

机器学习 · 计算机科学 2026-02-16 Atli Kosson , Jeremy Welborn , Yang Liu , Martin Jaggi , Xi Chen

Deep learning models have become a cornerstone of modern AI research, yet their initializations and learning rates may at times be set in an opaque or ad-hoc fashion due to the high cost of hyperparameter sweeps. The $\mu$-Parameterization…

机器学习 · 计算机科学 2025-02-17 Lucas Lingle

The Normalized Transformer, or nGPT (arXiv:2410.01131) achieves impressive training speedups and does not require weight decay or learning rate warmup. However, despite having hyperparameters that explicitly scale with model size, we…

机器学习 · 计算机科学 2026-05-04 Boris Shigida , Boris Hanin , Andrey Gromov

By classifying infinite-width neural networks and identifying the *optimal* limit, Tensor Programs IV and V demonstrated a universal way, called $\mu$P, for *widthwise hyperparameter transfer*, i.e., predicting optimal hyperparameters of…

神经与进化计算 · 计算机科学 2023-10-13 Greg Yang , Dingli Yu , Chen Zhu , Soufiane Hayou

Local learning, which trains a network through layer-wise local targets and losses, has been studied as an alternative to backpropagation (BP) in neural computation. However, its algorithms often become more complex or require additional…

机器学习 · 计算机科学 2025-05-22 Satoki Ishikawa , Rio Yokota , Ryo Karakida

Deeper modern architectures are costly to train, making hyperparameter transfer preferable to expensive repeated tuning. Maximal Update Parametrization ($\mu$P) helps explain why many hyperparameters transfer across width. Yet depth scaling…

机器学习 · 计算机科学 2026-02-10 Shenxi Wu , Haosong Zhang , Xingjian Ma , Shirui Bian , Yichi Zhang , Xi Chen , Wei Lin

Hyperparameter transfer allows extrapolating optimal optimization hyperparameters from small to large scales, making it critical for training large language models (LLMs). This is done either by fitting a scaling law to the hyperparameters…

机器学习 · 计算机科学 2026-05-21 Dayal Singh Kalra , Maissam Barkeshli

We establish four structural results for feature learning in wide two-layer neural networks under the Maximal Update Parametrization ($\mu$P). First, we prove global existence and uniqueness of the mean-field limit of noisy gradient descent…

机器学习 · 计算机科学 2026-05-26 Akmal Xodarev

In this short note we consider random fully connected ReLU networks of width $n$ and depth $L$ equipped with a mean-field weight initialization. Our purpose is to study the dependence on $n$ and $L$ of the maximal update ($\mu$P) learning…

机器学习 · 计算机科学 2023-05-16 Samy Jelassi , Boris Hanin , Ziwei Ji , Sashank J. Reddi , Srinadh Bhojanapalli , Sanjiv Kumar

Choosing an appropriate learning rate remains a key challenge in scaling depth of modern deep networks. The classical maximal update parameterization ($\mu$P) enforces a fixed per-layer update magnitude, which is well suited to homogeneous…

机器学习 · 计算机科学 2025-12-01 Haosong Zhang , Shenxi Wu , Yichi Zhang , Xi Chen , Wei Lin

Hyperparameter transfer across model architectures dramatically reduces the amount of compute necessary for tuning large language models (LLMs). The maximal update parameterization ({\mu}P) ensures transfer through principled mathematical…

One of the main challenges in optimal scaling of large language models (LLMs) is the prohibitive cost of hyperparameter tuning, particularly learning rate $\eta$ and batch size $B$. While techniques like $\mu$P (Yang et al., 2022) provide…

机器学习 · 计算机科学 2025-01-10 Oleg Filatov , Jan Ebert , Jiangtao Wang , Stefan Kesselheim

We consider a transfer-learning problem by using the parameter transfer approach, where a suitable parameter of feature mapping is learned through one task and applied to another objective task. Then, we introduce the notion of the local…

机器学习 · 统计学 2017-01-19 Wataru Kumagai

Transfer learning (TL) is a well-established machine learning technique to boost the generalization performance on a specific (target) task using information gained from a related (source) task, and it crucially depends on the ability of a…

无序系统与神经网络 · 物理学 2024-07-11 Alessandro Ingrosso , Rosalba Pacelli , Pietro Rotondo , Federica Gerace

The cost of hyperparameter tuning in deep learning has been rising with model sizes, prompting practitioners to find new tuning methods using a proxy of smaller networks. One such proposal uses $\mu$P parameterized networks, where the…

机器学习 · 统计学 2023-12-11 Blake Bordelon , Lorenzo Noci , Mufan Bill Li , Boris Hanin , Cengiz Pehlevan

Scaling limits, such as infinite-width limits, serve as promising theoretical tools to study large-scale models. However, it is widely believed that existing infinite-width theory does not faithfully explain the behavior of practical…

机器学习 · 计算机科学 2025-10-28 Moritz Haas , Sebastian Bordt , Ulrike von Luxburg , Leena Chennuru Vankadara

We study the fundamental problem of transfer learning where a learning algorithm collects data from some source distribution $P$ but needs to perform well with respect to a different target distribution $Q$. A standard change of measure…

机器学习 · 计算机科学 2024-03-19 Alkis Kalavasis , Ilias Zadik , Manolis Zampetakis

The growing scale of deep learning models has rendered standard hyperparameter (HP) optimization prohibitively expensive. A promising solution is the use of scale-aware hyperparameters, which can enable direct transfer of optimal HPs from…

机器学习 · 计算机科学 2025-12-30 Nikhil Ghosh , Denny Wu , Alberto Bietti

Despite deep neural networks' powerful representation learning capabilities, theoretical understanding of how networks can simultaneously achieve meaningful feature learning and global convergence remains elusive. Existing approaches like…

机器学习 · 计算机科学 2025-07-23 Zixiang Chen , Greg Yang , Qingyue Zhao , Quanquan Gu

Human learners have the natural ability to use knowledge gained in one setting for learning in a different but related setting. This ability to transfer knowledge from one task to another is essential for effective learning. In this paper,…

统计理论 · 数学 2019-06-10 T. Tony Cai , Hongji Wei
‹ 上一页 1 2 3 10 下一页 ›