English
Related papers

Related papers: Bug or Feature$^2$: Weight Drift, Activation Spars…

200 papers

We study neural networks with trainable low-degree rational activation functions and show that they are more expressive and parameter-efficient than modern piecewise-linear and smooth activations such as ELU, LeakyReLU, LogSigmoid, PReLU,…

Machine Learning · Computer Science 2026-02-16 Maosen Tang , Alex Townsend

The ability of neural networks to provide `best in class' approximation across a wide range of applications is well-documented. Nevertheless, the powerful expressivity of neural networks comes to naught if one is unable to effectively train…

Machine Learning · Computer Science 2020-07-15 Mark Ainsworth , Yeonjong Shin

Behavior of deep neural networks can be inconsistent between different versions. Regressions during model update are a common cause of concern that often over-weigh the benefits in accuracy or efficiency gain. This work focuses on…

Computation and Language · Computer Science 2021-05-10 Yuqing Xie , Yi-an Lai , Yuanjun Xiong , Yi Zhang , Stefano Soatto

Deep neural networks exhibit periodic loss spikes during unregularized long-term training, a phenomenon known as the "Slingshot Mechanism." Existing work usually attributes this to intrinsic optimization dynamics, but its triggering…

Machine Learning · Computer Science 2026-05-27 Liu Hanqing , Jianjun Cao , Yuanze Li , Zijian Zhou

In this paper, we first identify activation shift, a simple but remarkable phenomenon in a neural network in which the preactivation value of a neuron has non-zero mean that depends on the angle between the weight vector of the neuron and…

Neural and Evolutionary Computing · Computer Science 2024-03-22 Takuro Kutsuna

We analyze a simple one-hidden-layer neural network with ReLU activation functions and fixed biases, with one-dimensional input and output. We study both continuous and discrete versions of the model, and we rigorously prove the convergence…

Machine Learning · Computer Science 2026-04-10 Fabricio Macià , Shu Nakamura

Deep learning has been widely used in many fields, but the model training process usually consumes massive computational resources and time. Therefore, designing an efficient neural network training method with a provable convergence…

Machine Learning · Computer Science 2023-07-14 Lianke Qin , Zhao Song , Yuanyuan Yang

Activation functions are fundamental for enabling nonlinear representations in deep neural networks. However, the standard rectified linear unit (ReLU) often suffers from inactive or "dead" neurons caused by its hard zero cutoff. To address…

Machine Learning · Computer Science 2025-11-12 Md Motaleb Hossen Manik , Md Zabirul Islam , Ge Wang

Sparse Neural Networks (SNNs) can potentially demonstrate similar performance to their dense counterparts while saving significant energy and memory at inference. However, the accuracy drop incurred by SNNs, especially at high pruning…

Machine Learning · Computer Science 2023-06-06 Mohammad Loni , Aditya Mohan , Mehdi Asadi , Marius Lindauer

We propose ReDense as a simple and low complexity way to improve the performance of trained neural networks. We use a combination of random weights and rectified linear unit (ReLU) activation function to add a ReLU dense (ReDense) layer to…

Machine Learning · Computer Science 2020-10-27 Alireza M. Javid , Sandipan Das , Mikael Skoglund , Saikat Chatterjee

A wide variety of activation functions have been proposed for neural networks. The Rectified Linear Unit (ReLU) is especially popular today. There are many practical reasons that motivate the use of the ReLU. This paper provides new…

Machine Learning · Statistics 2020-10-19 Rahul Parhi , Robert D. Nowak

We prove several hardness results for training depth-2 neural networks with the ReLU activation function; these networks are simply weighted sums (that may include negative coefficients) of ReLUs. Our goal is to output a depth-2 neural…

Machine Learning · Computer Science 2020-11-30 Surbhi Goel , Adam Klivans , Pasin Manurangsi , Daniel Reichman

Pruning of deep neural networks has been an effective technique for reducing model size while preserving most of the performance of dense networks, crucial for deploying models on memory and power-constrained devices. While recent sparse…

Computer Vision and Pattern Recognition · Computer Science 2025-12-02 Andy Li , Aiden Durrant , Milan Markovic , Tianjin Huang , Souvik Kundu , Tianlong Chen , Lu Yin , Georgios Leontidis

In this paper we investigate the performance of different types of rectified activation functions in convolutional neural network: standard rectified linear unit (ReLU), leaky rectified linear unit (Leaky ReLU), parametric rectified linear…

Machine Learning · Computer Science 2015-11-30 Bing Xu , Naiyan Wang , Tianqi Chen , Mu Li

Stochastic neural net weights are used in a variety of contexts, including regularization, Bayesian neural nets, exploration in reinforcement learning, and evolution strategies. Unfortunately, due to the large number of weights, all the…

Machine Learning · Computer Science 2018-04-03 Yeming Wen , Paul Vicol , Jimmy Ba , Dustin Tran , Roger Grosse

Deep learning models are effective for sequential data modeling, yet commonly used activation functions such as ReLU, LeakyReLU, and PReLU often exhibit gradient instability when applied to noisy, non-stationary financial time series. This…

Machine Learning · Computer Science 2026-01-26 George Awiakye-Marfo , Elijah Agbosu , Victoria Mawuena Barns , Samuel Asante Gyamerah

We introduce stochastic activations. This novel strategy randomly selects between several non-linear functions in the feed-forward layer of a large language model. In particular, we choose between SILU or RELU depending on a Bernoulli draw.…

Sparsification of neural networks is one of the effective complexity reduction methods to improve efficiency and generalizability. We consider the problem of learning a one hidden layer convolutional neural network with ReLU activation…

Optimization and Control · Mathematics 2020-02-26 Thu Dinh , Jack Xin

Gradient-based neural network training traditionally enforces symmetry between forward and backward propagation, requiring activation functions to be differentiable (or sub-differentiable) and strictly monotonic in certain regions to…

Neural and Evolutionary Computing · Computer Science 2025-09-10 Luigi Troiano , Francesco Gissi , Vincenzo Benedetto , Genny Tortora

Modern neural networks, although achieving state-of-the-art results on many tasks, tend to have a large number of parameters, which increases training time and resource usage. This problem can be alleviated by pruning. Existing methods,…

Machine Learning · Computer Science 2020-09-08 Andrei Apostol , Maarten Stol , Patrick Forré