English
Related papers

Related papers: Subcritical Signal Propagation at Initialization i…

200 papers

Deep neural networks are notorious for defying theoretical treatment. However, when the number of parameters in each layer tends to infinity, the network function is a Gaussian process (GP) and quantitatively predictive description is…

Machine Learning · Computer Science 2023-10-09 Darshil Doshi , Tianyu He , Andrey Gromov

Finding the right initialisation for neural networks is crucial to ensure smooth training and good performance. In transformers, the wrong initialisation can lead to one of two failure modes of self-attention layers: rank collapse, where…

Machine Learning · Statistics 2026-05-22 Alessio Giorlandino , Sebastian Goldt

Deep neural networks (DNNs) are known to be vulnerable to adversarial examples that are crafted with imperceptible perturbations, i.e., a small change in an input image can induce a mis-classification, and thus threatens the reliability of…

Machine Learning · Computer Science 2022-11-15 Deyin Liu , Lin Wu , Haifeng Zhao , Farid Boussaid , Mohammed Bennamoun , Xianghua Xie

Despite powering modern AI, transformers remain mysteriously brittle to train. We develop a stability theory that explains why pre-LayerNorm works, why DeepNorm uses $N^{-1/4}$ scaling, and why warmup is necessary, all from first…

Machine Learning · Computer Science 2026-02-24 Seyed Morteza Emadi

Deep neural networks are known to suffer from exploding or vanishing gradients as depth increases, a phenomenon closely tied to the spectral behavior of the input-output Jacobian. Prior work has identified critical initialization schemes…

Machine Learning · Computer Science 2025-11-25 Benjamin Dadoun , Soufiane Hayou , Hanan Salam , Mohamed El Amine Seddik , Pierre Youssef

Skip connections and normalisation layers form two standard architectural components that are ubiquitous for the training of Deep Neural Networks (DNNs), but whose precise roles are poorly understood. Recent approaches such as Deep Kernel…

Machine Learning · Computer Science 2023-02-22 Bobby He , James Martens , Guodong Zhang , Aleksandar Botev , Andrew Brock , Samuel L Smith , Yee Whye Teh

Transient amplification has been proposed as an important mechanism not only in neuroscience but in many areas modeled by dynamical systems. Despite that, there is no clear biologically plausible mechanism which fine-tunes the coupling…

Disordered Systems and Neural Networks · Physics 2020-11-18 Wojciech Tarnowski

The generalization error of deep neural networks via their classification margin is studied in this work. Our approach is based on the Jacobian matrix of a deep neural network and can be applied to networks with arbitrary non-linearities…

Machine Learning · Statistics 2017-07-04 Jure Sokolic , Raja Giryes , Guillermo Sapiro , Miguel R. D. Rodrigues

While the expressive power and computational capabilities of graph neural networks (GNNs) have been theoretically studied, their optimization and learning dynamics, in general, remain largely unexplored. Our study undertakes the Graph…

Machine Learning · Computer Science 2023-10-26 Nimrah Mustafa , Aleksandar Bojchevski , Rebekka Burkholz

Deep neural networks have lately shown tremendous performance in various applications including vision and speech processing tasks. However, alongside their ability to perform these tasks with such high accuracy, it has been shown that they…

Machine Learning · Computer Science 2019-05-29 Daniel Jakubovitz , Raja Giryes

Training recurrent neural networks (RNNs) on long sequence tasks is plagued with difficulties arising from the exponential explosion or vanishing of signals as they propagate forward or backward through the network. Many techniques have…

Machine Learning · Computer Science 2019-05-27 Dar Gilboa , Bo Chang , Minmin Chen , Greg Yang , Samuel S. Schoenholz , Ed H. Chi , Jeffrey Pennington

While the authors of Batch Normalization (BN) identify and address an important problem involved in training deep networks-- Internal Covariate Shift-- the current solution has certain drawbacks. Specifically, BN depends on batch statistics…

Machine Learning · Statistics 2016-07-13 Devansh Arpit , Yingbo Zhou , Bhargava U. Kota , Venu Govindaraju

Good initialization is essential for training Deep Neural Networks (DNNs). Oftentimes such initialization is found through a trial and error approach, which has to be applied anew every time an architecture is substantially modified, or…

Machine Learning · Statistics 2022-06-29 Tianyu He , Darshil Doshi , Andrey Gromov

We present a theoretical analysis of the Jacobian of an attention block within a transformer, showing that it is governed by the query, key, and value projections that define the attention mechanism. Leveraging this insight, we introduce a…

Machine Learning · Computer Science 2026-03-10 Hemanth Saratchandran , Simon Lucey

In practice it is often found that large over-parameterized neural networks generalize better than their smaller counterparts, an observation that appears to conflict with classical notions of function complexity, which typically favor…

Machine Learning · Statistics 2018-06-20 Roman Novak , Yasaman Bahri , Daniel A. Abolafia , Jeffrey Pennington , Jascha Sohl-Dickstein

We address the problem of estimating statistics of hidden units in a neural network using a method of analytic moment propagation. These statistics are useful for approximate whitening of the inputs in front of saturating non-linearities…

Machine Learning · Computer Science 2018-03-29 Alexander Shekhovtsov , Boris Flach

Normalization layers are one of the key building blocks for deep neural networks. Several theoretical studies have shown that batch normalization improves the signal propagation, by avoiding the representations from becoming collinear…

Machine Learning · Computer Science 2023-10-04 Alexandru Meterez , Amir Joudaki , Francesco Orabona , Alexander Immer , Gunnar Rätsch , Hadi Daneshmand

Transformer-based models have demonstrated exceptional performance across diverse domains, becoming the state-of-the-art solution for addressing sequential machine learning problems. Even though we have a general understanding of the…

Disordered Systems and Neural Networks · Physics 2024-06-12 Ángel Poc-López , Miguel Aguilera

Recent work has shown that tight concentration of the entire spectrum of singular values of a deep network's input-output Jacobian around one at initialization can speed up learning by orders of magnitude. Therefore, to guide important…

Machine Learning · Statistics 2018-02-28 Jeffrey Pennington , Samuel S. Schoenholz , Surya Ganguli

Genes are connected in regulatory networks, often modelled by ordinary differential equations. Changes in expression of a gene propagate to other genes along paths in the network. At a stable state, the system's Jacobian matrix confers…

Molecular Networks · Quantitative Biology 2014-11-03 Arne B. Gjuvsland , Erik Plahte
‹ Prev 1 2 3 10 Next ›