The paper explores how ReLU DNNs can represent MPC policies and vice versa.
problem Representing MPC policies as ReLU DNNs and vice versa.
method Developed an approximate method for identifying input-space in ReLU nets resulting in PWA functions over polyhedral regions. Studied inverse multiparametric linear or quadratic programs for reconstruction of constraints and cost functions given a PWA function.
result Identification and representation of MPC policies as ReLU DNNs and vice versa.
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…
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. Abstract Neural Networks (ANNs) improve DNN verification efficiency.
problem Efficiently verify safety-critical DNNs without slowing exponentially.
method Introduces ANNs that use abstract domains and activation functions to overapproximate DNNs.
result ANNs can soundly overapproximate DNNs with fewer nodes, improving verification efficiency.
Smooth DNNs mitigate the curse of dimensionality in uniform convergence for various regression tasks.
problem The curse of dimensionality in uniform convergence of ReLU networks.
method Analysis of smoothly activated deep neural networks (smooth DNNs), establishing pseudo-dimension bounds and non-asymptotic approximation guarantees.
result Smooth DNNs achieve non-asymptotic uniform convergence rates across multiple statistical contexts, mitigating the curse of dimensionality.
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.
Mean field theory has been successfully used to analyze deep neural networks (DNN) in the infinite size limit. Given the finite size of realistic DNN, we utilize the large deviation theory and path integral analysis to study the deviation of functions represented by DNN from their typical mean field solutions. The para…
Enhances DNN robustness and accuracy with L2,∞ normalization.
problem Improving the robustness and accuracy of deep neural networks.
method Introducing L2,∞ normalization of weight matrices in DNNs with Relu activation. result Lower bound for robustness measure in terms of L2,∞ norm and upper bound for Rademacher complexity. Deep neural networks approximate option prices in high-dimensional Lévy models efficiently.
problem Approximating option prices in high-dimensional financial models with jumps.
method Use of deep ReLU neural networks to approximate option prices in multivariate Lévy processes with polynomial growth in network size and dimension.
result Established sufficient conditions for polynomial growth in network size and dimension to approximate option prices with error ε.
Paper reveals hidden convexities in deep learning models using sparse signal processing.
problem Non-convex loss functions in deep learning models complicate optimization and theoretical understanding.
method Developed convex equivalences of ReLU NNs and their connections to sparse signal processing models.
result Recent research has uncovered hidden convexities in certain NN architectures, notably two-layer ReLU networks and other architectures.
Empirical studies show that gradient-based methods can learn deep neural networks (DNNs) with very good generalization performance in the over-parameterization regime, where DNNs can easily fit a random labeling of the training data. Very recently, a line of work explains in theory that with over-parameterization and p…
This paper simplifies deep ReLU networks into local linear models for better interpretability.
problem Limited transparency and interpretability of deep neural networks, especially ReLU networks.
method Local linear representation and equivalent set of local linear models (LLMs).
result Simplified deep ReLU networks for better interpretability and diagnostics.
The paper analyzes the role of ReLU gates in deep learning networks.
problem Understanding the role of gates in deep learning networks.
method Developed neural path features (NPF) and neural path values (NPV) to characterize the active sub-networks during training.
result The neural path kernel associated with NPFs is a fundamental quantity that characterizes the information stored in the gates of a DNN.
New proof shows efficient ReLU networks for piecewise linear functions.
problem Existence of efficient ReLU neural networks for piecewise linear functions.
method Degree 1 triangulations of the relative homology class bounded by polyhedra.
result Existence of efficient ReLU neural networks for functions with compact support.
AutoInit automatically finds good neural network initialization.
problem Finding optimal neural network initialization is crucial but time-consuming.
method Uses Jacobian tuning to automatically adjust network hyperparameters.
result The method finds good initialization for various network architectures.
Study on the complexity of 1D ReLU neural networks, proving growth in linear regions.
problem Understanding the complexity and expressivity of 1D ReLU neural networks.
method Analyzing the number of linear regions in randomly initialized, fully connected 1D ReLU networks in the infinite-width limit.
result The expected number of linear regions grows as a function of the number of neurons in each layer.
New neural network architectures use signed permutation representations for finite groups, improving performance.
problem Designing and optimizing neural networks for finite groups with signed permutation representations.
method Introduces G-invariant deep neural networks with densely connected layers and signed permutation representations. result Signed permutation representations lead to significantly better performance in classification tasks.
Studying the invertibility of deep neural networks (DNNs) provides a principled approach to better understand the behavior of these powerful models. Despite being a promising diagnostic tool, a consistent theory on their invertibility is still lacking. We derive a theoretically motivated approach to explore the preimag…
Deep neural networks (DNNs) have emerged as key enablers of machine learning. Applying larger DNNs to more diverse applications is an important challenge. The computations performed during DNN training and inference are dominated by operations on the weight matrices describing the DNN. As DNNs incorporate more layers a…
Improved generalization bounds for CNNs using Rademacher complexity.
problem Establishing non-vacuous generalization bounds for deep learning models.
method Rademacher complexity framework with novel contraction lemmas for high-dimensional mappings.
result Enhanced generalization bounds for a broader class of activation functions.
The paper examines when NTK theory applies to real finite-width neural networks.
problem Understanding when NTK theory accurately predicts the behavior of finite-width neural networks.
method Empirical study of fully-connected ReLU and sigmoid DNNs with various hyperparameters and depths.
result NTK theory does not always apply to sufficiently deep networks with exploding gradients, and the kernel changes significantly during training.
Deep neural networks (DNNs) generate much richer function spaces than shallow networks. Since the function spaces induced by shallow networks have several approximation theoretic drawbacks, this explains, however, not necessarily the success of deep networks. In this article we take another route by comparing the expre…
SGD methods fail to converge to global minimizers in deep neural networks with ReLU activation.
problem Failure of SGD methods to converge to global minimizers in deep neural networks.
method Stochastic Gradient Descent (SGD) and its variants like Adam, RMSProp, etc.
result SGD methods fail to converge to global minimizers with high probability in deep neural networks with ReLU activation.
Novel oracle-type inequality for logistic loss in DNNs achieves sharp convergence rates.
problem Generalization analysis for binary classification with DNNs and logistic loss.
method Established an oracle-type inequality to handle the boundedness of the target function.
result Optimal convergence rates for fully connected ReLU DNN classifiers trained with logistic loss.
We theoretically discuss why deep neural networks (DNNs) performs better than other models in some cases by investigating statistical properties of DNNs for non-smooth functions. While DNNs have empirically shown higher performance than other standard methods, understanding its mechanism is still a challenging problem.…
By lifting the ReLU function into a higher dimensional space, we develop a smooth multi-convex formulation for training feed-forward deep neural networks (DNNs). This allows us to develop a block coordinate descent (BCD) training algorithm consisting of a sequence of numerically well-behaved convex optimizations. Using…
Paper uses DNN-based MoM-GAN to estimate contaminated data distributions.
problem Estimating distributions of data with outliers or contamination.
method Combines GAN and MoM estimation with DNN for modeling.
result DNN-based MoM-GAN achieves better error bounds than other methods.
Improved DNN estimator with scalable subsampling for efficient inference.
problem Efficient inference for deep neural networks (DNNs).
method Non-random subsampling technique (scalable subsampling) applied to DNNs.
result Subagged DNN estimator offers computational efficiency and accurate point estimation/prediction intervals.
Deep neural networks achieve optimal classification rates in high dimensions.
problem Binary classification on high-dimensional data with specific smoothness and composition properties.
method Proved optimal convergence rate for ReLU DNNs trained with hinge loss.
result ReLU DNNs achieve optimal classification rates up to a logarithmic factor.
LQF linearizes deep models for better interpretability.
problem Lack of interpretability in deep neural networks.
method Simple modifications to architecture, loss function, and optimization.
result Comparable performance to non-linear fine-tuning, with interpretability.
Deep nets solve MDPs without high dimensions.
problem Solving Bellman equations for MDPs in high dimensions.
method Deep neural networks with ReLU activation approximating payoff and transition functions.
result Deep nets can approximate Q-functions in polynomially bounded parameters. In recent years, deep neural networks (DNNs) have been applied to various machine leaning tasks, including image recognition, speech recognition, and machine translation. However, large DNN models are needed to achieve state-of-the-art performance, exceeding the capabilities of edge devices. Model reduction is thus nee…
Deep neural networks (DNN) trained in a supervised way suffer from two known problems. First, the minima of the objective function used in learning correspond to data points (also known as rubbish examples or fooling images) that lack semantic similarity with the training data. Second, a clean input can be changed by a…
We derive the fast convergence rates of a deep neural network (DNN) classifier with the rectified linear unit (ReLU) activation function learned using the hinge loss. We consider three cases for a true model: (1) a smooth decision boundary, (2) smooth conditional class probability, and (3) the margin condition (i.e., t…
A bias classifier is introduced to resist adversarial attacks.
problem Resisting adversarial attacks on deep neural networks (DNNs).
method Introducing the bias part of a DNN with Relu as the activation function as a classifier, and adding a random first-degree part to make it information-theoretically safe.
result The bias classifier is more robust than DNNs of similar size against adversarial attacks.
Activation functions influence behavior and performance of DNNs. Nonlinear activation functions, like Rectified Linear Units (ReLU), Exponential Linear Units (ELU) and Scaled Exponential Linear Units (SELU), outperform the linear counterparts. However, selecting an appropriate activation function is a challenging probl…
Derives moments of PL networks for robust DNNs.
problem Analyzing robustness of DNNs to adversarial attacks.
method Derives exact analytic expressions for moments of PL networks, generalizes variance expression, and constructs sparse and smooth adversarial attacks.
result New variance expressions are tighter and can be efficiently approximated.
We analyze deep neural networks using convex duality to reveal hidden layer structures.
problem Understanding the structure of deep neural networks.
method Introducing a convex analytic framework to characterize hidden layer weights.
result Optimal hidden layer weights align with previous layers via duality.
Understanding the representational power of Deep Neural Networks (DNNs) and how their structural properties (e.g., depth, width, type of activation unit) affect the functions they can compute, has been an important yet challenging question in deep learning and approximation theory. In a seminal paper, Telgarsky highlig…
Deep learning transforms data geometrically, akin to Ricci flow, improving classification accuracy.
problem Understanding geometric transformations in non-smooth activation functions.
method Developed a computational framework to quantify geometric changes in DNNs and introduced the concept of `global Ricci network flow`.
result Global Ricci network flow correlates with DNN accuracy, independent of network architecture and data set.
Initializing the weights and the biases is a key part of the training process of a neural network. Unlike the subsequent optimization phase, however, the initialization phase has gained only limited attention in the literature. In this paper we discuss some consequences of commonly used initialization strategies for va…
It remains a puzzle that why deep neural networks (DNNs), with more parameters than samples, often generalize well. An attempt of understanding this puzzle is to discover implicit biases underlying the training process of DNNs, such as the Frequency Principle (F-Principle), i.e., DNNs often fit target functions from lo…
This paper presents a new optimisation approach to train Deep Neural Networks (DNNs) with discriminative sequence criteria. At each iteration, the method combines information from the Natural Gradient (NG) direction with local curvature information of the error surface that enables better paths on the parameter manifol…
We analyze architectural features of Deep Neural Networks (DNNs) using the so-called Neural Tangent Kernel (NTK), which describes the training and generalization of DNNs in the infinite-width setting. In this setting, we show that for fully-connected DNNs, as the depth grows, two regimes appear: "order", where the (sca…
Deep neural networks (DNNs) have demonstrated dominating performance in many fields; since AlexNet, networks used in practice are going wider and deeper. On the theoretical side, a long line of works has been focusing on training neural networks with one hidden layer. The theory of multi-layer networks remains largely …
New initialization schemes preserve fractional moments of weights in deep networks, improving training and test performance.
problem Heavy-tailed distribution of stochastic gradients in DNNs during training.
method Developed initialization schemes that preserve any given fractional moment of order s < 2 over layers for various activations.
result The network output admits a heavy-tailed distribution with finite moments, improving training and test performance.
Post-shifted BN prevents filter collapse in BN networks, improving model performance.
problem Filter collapse in BN networks reduces network capacity and harms model performance.
method Post-shifted BN (psBN) to prevent filter collapse by making BN parameters trainable again.
result psBN prevents filter collapse and increases model performance in various tasks.
This work explains neural collapse in shallow neural networks and its impact on generalization.
problem Understanding neural collapse in shallow neural networks and its effect on generalization.
method Analysis of two and three-layer ReLU neural networks, focusing on data dimension, sample size, and signal-to-noise ratio.
result Neural collapse occurs in shallow ReLU networks under certain conditions related to data properties and network architecture.