ReLU units can 'die' in neural networks, causing slower convergence.
problem ReLU units sometimes produce near-zero outputs during training.
method Simulation and statistical analysis of a simplified ReLU unit model.
result Activation probability decreases as training progresses, leading to slower convergence.
A new method for optimizing regression problems with ReLU units converges.
problem Optimizing regression problems involving ReLU units in large language models.
method Introduced a greedy algorithm based on approximate Newton method, proving convergence in terms of the distance to optimal solution.
result The method converges in the sense of the distance to optimal solution under certain assumptions.
ReLU activations lead to smoother learning curves compared to sigmoidal activations in neural networks.
problem Comparing the performance of ReLU and sigmoidal activations in neural networks.
method Analytical computation of learning curves in shallow networks with different activation functions.
result ReLU networks exhibit continuous transitions in performance, while sigmoidal networks show discontinuous transitions.
The paper investigates how target normalization and momentum affect dying ReLUs in neural networks.
problem Understanding and mitigating the dying ReLU problem in neural networks.
method Empirical analysis and theoretical modeling of a discrete-time linear autonomous system.
result Target variance plays a crucial role in the dying ReLU phenomenon, and momentum exacerbates this issue.
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 unit (PReLU) and a new randomized leaky rectified linear units (RReLU). We evalu…
Paper interprets deep learning using decision trees and Haar wavelets.
problem Understanding the function approximation capabilities of ReLU deep learning.
method Constructing a deep learning structure equivalent to a forest and approximating Haar wavelet functions with ReLU deep learning.
result ReLU deep learning can be considered as decision trees and approximates Haar wavelet functions with arbitrary precision.
Neural networks approximate unit spheres as polytopes.
problem Approximating unit spheres with neural networks.
method Using ReLU activation in neural networks to generate polytopes.
result Neural networks can approximate unit spheres as polytopes.
PHP connects to ReLU neural networks for scalable Bayesian inference.
problem Scalability and Bayesian inference in two-layer ReLU neural networks.
method PHP with Gaussian prior, decomposition propositions, annealed sequential Monte Carlo.
result PHP provides an alternative scalable representation for two-layer ReLU neural networks.
Study approximates nonlinear functionals using deep ReLU networks.
problem Approximating nonlinear continuous functionals with neural networks.
method Constructs continuous piecewise linear interpolation under simple triangulation, analyzes rates of approximation.
result Established rates of approximation for functional deep ReLU networks.
This paper explores transforming deep neural networks into simpler structures while preserving or approximating their functionality.
problem Transforming deep neural networks into simpler structures while preserving or approximating their functionality.
method Constructive proof of global linear approximation, removal of inactive or always active ReLUs, and experiments with regularization and adversarial training.
result For any feed-forward ReLU network, there exists a global linear approximation to a 2-hidden-layer shallow network with a fixed number of units.
Improved activation function NLReLU boosts neural network performance.
problem Performance issues with ReLU activation function.
method NLReLU uses parametric natural logarithmic transform to improve ReLU.
result NLReLU provides higher accuracy than ReLU in various neural networks.
Visualizes decision boundaries of neural networks using geometric illustrations.
problem Understanding decision boundaries in neural networks during training.
method Developed GINN tool to geometrically illustrate ReLU units' activation states.
result Observed and discussed several phenomena during neural network training.
New neural network units resist adversarial attacks effectively.
problem Adversarial attacks on machine learning models.
method Introduced MWD units, developed training techniques, and computed robustness.
result MWD networks are significantly more robust to adversarial attacks.
Study shows shallow ReLU networks struggle with high-dimensional Lipschitz functions.
problem Expressing high-dimensional Lipschitz functions with shallow ReLU networks.
method Established lower bounds on shallow network complexity for polynomial approximation.
result Shallow ReLU networks suffer from the curse of dimensionality for Lipschitz functions.
We study deep Bayesian neural networks with Gaussian priors, revealing heavy-tailed unit activations.
problem Characterizing regularization effects in deep Bayesian neural networks.
method Investigation of deep Bayesian neural networks with Gaussian weight priors and ReLU-like nonlinearities.
result The prior distribution on units becomes increasingly heavy-tailed with depth, influencing activation patterns.
Algorithm learns two-layer residual units using ReLU activations from samples.
problem Learning two-layer residual units from samples.
method Design layer-wise objectives as functionals, formulate ERM as QP, solve using LP, prove statistical consistency.
result Strong statistical consistency and robustness of the algorithm.
SPLASH units improve deep neural network accuracy and adversarial robustness.
problem Improving deep neural network accuracy and adversarial robustness.
method SPLASH units are learnable activation functions with specific properties that enhance both accuracy and robustness.
result SPLASH units significantly increase adversarial robustness of deep neural networks.
In adversarial attacks to machine-learning classifiers, small perturbations are added to input that is correctly classified. The perturbations yield adversarial examples, which are virtually indistinguishable from the unperturbed input, and yet are misclassified. In standard neural networks used for deep learning, atta…
Analyzes how ReLU affects neural network capacity and solution space geometry.
problem Understanding the effects of ReLU on neural network capacity and solution space.
method Analytical and numerical studies of two-layer neural networks with ReLU activations.
result The capacity of ReLU networks remains finite as the number of hidden neurons increases, unlike for threshold units.
New Gated Linear Unit (GaLU) improves neural network performance.
problem Theoretical gap between ReLU networks and empirical observations in shallow networks.
method Introduced Gated Linear Unit (GaLU) by decoupling linearity from gating.
result Improved memorization and generalization results for GaLU networks.
The paper calculates bounds on the local Lipschitz constants of neural network layers.
problem Understanding the Lipschitz constants of neural network layers for robustness analysis.
method Analytical approach to determine upper bounds on local Lipschitz constants of affine-ReLU functions.
result The method produces tighter bounds than the standard conservative bound, especially for small perturbations.
Quantile regression with ReLU networks achieves minimax rates for various function types.
problem Estimating quantiles from covariates with neural networks.
method Quantile regression with rectified linear unit (ReLU) neural networks.
result ReLU networks achieve minimax rates for broad collections of function types.
In this paper we investigate the family of functions representable by deep neural networks (DNN) with rectified linear units (ReLU). We give an algorithm to train a ReLU DNN with one hidden layer to *global optimality* with runtime polynomial in the data size albeit exponential in the input dimension. Further, we impro…
Study on optimal ReLU networks with weight decay for interpolation.
problem Interpolating data with radially symmetric distributions using shallow ReLU networks.
method Weight decay regularization in infinite neuron, infinite data limit; analysis of growth rates.
result Existence and growth rates of unique radially symmetric minimizers with weight decay.
New activations improve deep network reproducibility without sacrificing accuracy.
problem Deep networks' reproducibility issues, especially on distributed systems.
method Developed SmeLU activations, smoother than ReLU, to enhance reproducibility.
result SmeLU activations provide better accuracy-reproducibility tradeoffs.
The Rectified Linear Unit (ReLU) is a foundational activation function in artficial neural networks. Recent literature frequently misattributes its origin to the 2018 (initial) version of this paper, which exclusively investigated ReLU at the classification layer. This paper formally corrects the citation record by tra…
Infinite width ReLU networks can approximate functions with bounded Euclidean norm.
problem Functions that can be approximated by ReLU networks with bounded Euclidean norm.
method Analyzing the minimal network norm required to approximate a given function.
result The minimal network norm for representing a function \( f \) is \( \max(\int |f''(x)| dx, |f'(-\infty) + f'(+\infty)|) \).
Researchers use statistical physics to model neural network learning dynamics.
problem Understanding the learning dynamics of ReLU neural networks.
method Developed a system of differential equations using statistical physics techniques.
result ReLU networks exhibit distinct learning behavior compared to sigmoidal networks.
This article concerns the expressive power of depth in neural nets with ReLU activations and bounded width. We are particularly interested in the following questions: what is the minimal width wmin(d) so that ReLU nets of width wmin(d) (and arbitrary depth) can approximate any continuous functio…
L*ReLU improves deep learning for fine-grained image classification.
problem Fine-grained image classification requires specific AFs.
method Proposes L*ReLU, piece-wise linear AFs for deep learning.
result L*ReLU achieves superior results on FGVC datasets.
Examines learning efficiency in neural networks and related models.
problem Analyzing efficiency in deep learning models with singular learning coefficients.
method Examined learning coefficients in neural networks and three-layer neural networks with ReLU units.
result Extended results to include Softmax function, providing a broader understanding of learning efficiency.
We present a novel neural network algorithm, the Tensor Switching (TS) network, which generalizes the Rectified Linear Unit (ReLU) nonlinearity to tensor-valued hidden units. The TS network copies its entire input vector to different locations in an expanded representation, with the location determined by its hidden un…
Algorithm removes units and layers of neural networks without losing accuracy.
problem Deploying large neural networks under limited resources.
method Mixed-Integer Linear Programming (MILP) and L1 regularization.
result Lossless compression of neural networks is achieved.
Deep ReLU networks need Ω(N) parameters to interpolate at irregularly spaced points.
problem Interpolating at irregularly spaced data points with deep ReLU networks.
method Analyzing the number of parameters required for interpolation.
result Ω(N) parameters are necessary for interpolation when δ is exponentially small in N.
We give the first dimension-efficient algorithms for learning Rectified Linear Units (ReLUs), which are functions of the form x↦max(0,w⋅x) with w∈Sn−1. Our algorithm works in the challenging Reliable Agnostic learning model of Kalai, Kanade, and Ma…
New maximum score estimators using ReLU functions and deep neural networks.
problem Estimating parameters in models with sign restrictions.
method ReLU-based maximum score criterion and DNN architecture.
result RMS estimator achieves n−s/(2s+1) convergence rate and asymptotic normality. Paper improves confidence intervals and variance estimation for deep learning models.
problem Improving confidence intervals and variance estimation in deep learning models.
method Residual-based framework for conditional variance estimation; robust bootstrap procedure for confidence intervals.
result First non-asymptotic bounds for variance estimation using ReLU networks.
Study on ReLU regression with Massart noise, achieving exact parameter recovery.
problem Efficiently fitting ReLUs to data in the presence of Massart noise.
method Developed an efficient algorithm for exact parameter recovery under mild assumptions.
result Achieved exact parameter recovery in ReLU regression with Massart noise.
Deep gated networks help understand training and generalization in deep learning.
problem Understanding the role of SGD in training and generalization of deep neural networks with ReLU activation.
method Developed deep gated networks (DGNs) as a framework to analyze training and generalization in DNNs with ReLU activation.
result Gate adaptation is key for generalization in deep neural networks.
`Biologically inspired' activation functions, such as the logistic sigmoid, have been instrumental in the historical advancement of machine learning. However in the field of deep learning, they have been largely displaced by rectified linear units (ReLU) or similar functions, such as its exponential linear unit (ELU) v…
Gradient descent optimizes deep ReLU networks with proper initialization.
problem Training deep neural networks with ReLU activation.
method Gradient descent and stochastic gradient descent with proper random weight initialization.
result Gradient descent finds global minima for over-parameterized deep ReLU networks.
Dropout training improves neural networks' performance.
problem Improving neural network convergence and generalization.
method Two-layer neural networks with ReLU activations, overparametrization, and positive margin assumption.
result Dropout training achieves ε-suboptimality in test error in O(1/ε) iterations.
We establish L∞ and L2 error bounds for functions of many variables that are approximated by linear combinations of ReLU (rectified linear unit) and squared ReLU ridge functions with ℓ1 and ℓ0 controls on their inner and outer parameters. With the squared ReLU ridge function, we show th…
New algorithms accelerate solving nonlinear matrix decomposition with ReLU.
problem Nonlinear matrix decomposition with ReLU function.
method Two new algorithms: A-NMD and 3B-NMD, with adaptive extrapolation and block parametrization.
result Effective algorithms accelerate solving ReLU-NMD problems.
The paper reveals how clustering in ReLU networks enhances interpretability.
problem Improving interpretability of deep learning models.
method Clustering of ReLU neuron patterns to reveal underlying affine maps.
result The network's predictions can be explained by feature importance within clusters.
Deep ReLU networks can approximate smooth functions nearly optimally.
problem Approximating smooth functions with deep neural networks.
method Using Taylor expansions and deep ReLU network approximations, the paper establishes optimal approximation error bounds.
result Deep ReLU networks of width and depth O(NlnN) and O(LlnL) can approximate f∈Cs([0,1]d) with an error O(∥f∥Cs([0,1]d)N−2s/dL−2s/d). ReLU networks don't exponentially distort curve lengths as previously thought.
problem Understanding how neural networks distort curve lengths with depth.
method Analyzing expected length distortion of ReLU networks with random initialization.
result Expected length distortion does not grow with depth, and shrinks slightly.
The paper analyzes why ReLU and related functions are effective in neural networks.
problem Understanding why certain activation functions are effective in neural networks.
method Using spline theory, the paper provides theoretical characterizations of activation functions.
result The paper explains the importance of activation functions and related strategies in neural network design.