English
Related papers

Related papers: Depthwise Hyperparameter Transfer in Residual Netw…

200 papers

We study the high-dimensional training dynamics of a shallow neural network with quadratic activation in a teacher-student setup. We focus on the extensive-width regime, where the teacher and student network widths scale proportionally with…

Optimization and Control · Mathematics 2026-01-16 Simon Martin , Giulio Biroli , Francis Bach

Deep neural networks' remarkable ability to correctly fit training data when optimized by gradient-based algorithms is yet to be fully understood. Recent theoretical results explain the convergence for ReLU networks that are wider than…

Machine Learning · Computer Science 2021-02-09 Asaf Noy , Yi Xu , Yonathan Aflalo , Lihi Zelnik-Manor , Rong Jin

Deep neural networks are powerful tools for solving nonlinear problems in science and engineering, but training highly accurate models becomes challenging as problem complexity increases. Non-convex optimization and sensitivity to…

Machine Learning · Computer Science 2026-04-20 Ethan Mulle , Wei Kang , Qi Gong

Scaling factors in residual branches have emerged as a prevalent method for boosting neural network performance, especially in normalization-free architectures. While prior work has primarily examined scaling effects from an optimization…

Machine Learning · Computer Science 2026-05-26 Zixiong Yu , Guhan Chen , Jianfa Lai , Bohan Li , Songtao Tian

In this note, we first derive a one-parameter family of hyperparameter scaling strategies that interpolates between the neural-tangent scaling and mean-field/maximal-update scaling. We then calculate the scalings of dynamical observables --…

Machine Learning · Computer Science 2022-10-19 Sho Yaida

Training deep neural networks with stochastic gradient descent (SGD) can often achieve zero training loss on real-world tasks although the optimization landscape is known to be highly non-convex. To understand the success of SGD for…

Machine Learning · Statistics 2020-06-15 Yiping Lu , Chao Ma , Yulong Lu , Jianfeng Lu , Lexing Ying

Deep neural networks (DNNs) are so over-parametrized that recent research has found them to already contain a subnetwork with high accuracy at their randomly initialized state. Finding these subnetworks is a viable alternative training…

Computer Vision and Pattern Recognition · Computer Science 2021-11-25 Ángel López García-Arias , Masanori Hashimoto , Masato Motomura , Jaehoon Yu

Training deep neural networks (DNNs) is computationally expensive, which is problematic especially when performing duplicated or similar training runs in model ensemble or fine-tuning pre-trained models, for example. Once we have trained…

Machine Learning · Computer Science 2023-10-04 Daiki Chijiwa

A residual network (or ResNet) is a standard deep neural net architecture, with state-of-the-art performance across numerous applications. The main premise of ResNets is that they allow the training of each layer to focus on fitting just…

Machine Learning · Computer Science 2018-09-28 Ohad Shamir

One of the arguments to explain the success of deep learning is the powerful approximation capacity of deep neural networks. Such capacity is generally accompanied by the explosive growth of the number of parameters, which, in turn, leads…

Machine Learning · Computer Science 2022-09-15 Zuowei Shen , Haizhao Yang , Shijun Zhang

Network pruning reduces the computation costs of an over-parameterized network without performance damage. Prevailing pruning algorithms pre-define the width and depth of the pruned networks, and then transfer parameters from the unpruned…

Computer Vision and Pattern Recognition · Computer Science 2019-10-17 Xuanyi Dong , Yi Yang

Parameter transfer is a central paradigm in transfer learning, enabling knowledge reuse across tasks and domains by sharing model parameters between upstream and downstream models. However, when only a subset of parameters from the upstream…

Machine Learning · Computer Science 2026-01-08 Hua Yuan , Xuran Meng , Qiufeng Wang , Shiyu Xia , Ning Xu , Xu Yang , Jing Wang , Xin Geng , Yong Rui

Deep learning methods are known to generalize well from training to future data, even in an overparametrized regime, where they could easily overfit. One explanation for this phenomenon is that even when their *ambient dimensionality*,…

Machine Learning · Computer Science 2025-05-22 Hossein Zakerinia , Dorsa Ghobadi , Christoph H. Lampert

Residual Network (ResNet) is the state-of-the-art architecture that realizes successful training of really deep neural network. It is also known that good weight initialization of neural network avoids problem of vanishing/exploding…

Machine Learning · Computer Science 2017-10-16 Masato Taki

Given an existing trained neural network, it is often desirable to learn new capabilities without hindering performance of those already learned. Existing approaches either learn sub-optimal solutions, require joint training, or incur a…

Computer Vision and Pattern Recognition · Computer Science 2018-02-15 Amir Rosenfeld , John K. Tsotsos

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…

Machine Learning · Computer Science 2025-12-01 Haosong Zhang , Shenxi Wu , 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…

Machine Learning · Computer Science 2026-05-21 Dayal Singh Kalra , Maissam Barkeshli

Despite achieving state-of-the-art performance, deep learning methods generally require a large amount of labeled data during training and may suffer from overfitting when the sample size is small. To ensure good generalizability of deep…

Computer Vision and Pattern Recognition · Computer Science 2020-06-30 Xiaoxu Li , Liyun Yu , Xiaochen Yang , Zhanyu Ma , Jing-Hao Xue , Jie Cao , Jun Guo

Very deep convolutional neural networks (CNNs) yield state of the art results on a wide variety of visual recognition problems. A number of state of the the art methods for image recognition are based on networks with well over 100 layers…

Computer Vision and Pattern Recognition · Computer Science 2016-07-15 Joel Moniz , Christopher Pal

Over-parameterization and adaptive methods have played a crucial role in the success of deep learning in the last decade. The widespread use of over-parameterization has forced us to rethink generalization by bringing forth new phenomena,…

Machine Learning · Statistics 2020-12-01 Vatsal Shah , Soumya Basu , Anastasios Kyrillidis , Sujay Sanghavi