The Singular Values of Convolutional Layers

2021

We study regularized deep neural networks (DNNs) and introduce a convex analytic framework to characterize the structure of the hidden layers. We show that a set of optimal hidden layer weights for a norm regularized DNN training problem can be explicitly found as the extreme points of a convex set. For the special case of deep linear networks with $K$ outputs, we prove that each optimal weight matrix is rank-$K$ and aligns with the previous layers via duality. More importantly, we apply the same characterization to deep ReLU networks with whitened data and prove the same weight alignment holds. As a corollary, we prove that norm regularized deep ReLU networks yield spline interpolation for one-dimensional datasets which was previously known only for two-layer networks. Furthermore, we provide closed-form solutions for the optimal layer weights when data is rank-one or whitened. We then verify our theory via numerical experiments.

arXiv (Cornell University), 2020

We study training of Convolutional Neural Networks (CNNs) with ReLU activations and introduce exact convex optimization formulations with a polynomial complexity with respect to the number of data samples, the number of neurons, and data dimension. More specifically, we develop a convex analytic framework utilizing semi-infinite duality to obtain equivalent convex optimization problems for several two-and three-layer CNN architectures. We first prove that two-layer CNNs can be globally optimized via an 2 norm regularized convex program. We then show that multi-layer circular CNN training problems with a single ReLU layer are equivalent to an 1 regularized convex program that encourages sparsity in the spectral domain. We also extend these results to three-layer CNNs with two ReLU layers. Furthermore, we present extensions of our approach to different pooling methods, which elucidates the implicit architectural bias as convex regularizers.

Applied and Computational Harmonic Analysis

Deep learning has been widely applied and brought breakthroughs in speech recognition, computer vision, and many other domains. Deep neural network architectures and computational issues have been well studied in machine learning. But there lacks a theoretical foundation for understanding the approximation or generalization ability of deep learning methods generated by the network architectures such as deep convolutional neural networks. Here we show that a deep convolutional neural network (CNN) is universal, meaning that it can be used to approximate any continuous function to an arbitrary accuracy when the depth of the neural network is large enough. This answers an open question in learning theory. Our quantitative estimate, given tightly in terms of the number of free parameters to be computed, verifies the efficiency of deep CNNs in dealing with large dimensional data. Our study also demonstrates the role of convolutions in deep CNNs.

2019

We prove bounds on the generalization error of convolutional networks. The bounds are in terms of the training loss, the number of parameters, the Lipschitz constant of the loss and the distance from the weights to the initial weights. They are independent of the number of pixels in the input, and the height and width of hidden feature maps. We present experiments with CIFAR-10, along with varying hyperparameters of a deep convolutional network, comparing our bounds with practical generalization gaps.

arXiv (Cornell University), 2020

We study regularized deep neural networks and introduce an analytic framework to characterize the structure of the hidden layers. We show that a set of optimal hidden layer weight matrices for a norm regularized deep neural network training problem can be explicitly found as the extreme points of a convex set. For two-layer linear networks, we first formulate a convex dual program and prove that strong duality holds. We then extend our derivations to prove that strong duality also holds for certain deep networks. In particular, for linear deep networks, we show that each optimal layer weight matrix is rank-one and aligns with the previous layers when the network output is scalar. We also extend our analysis to the vector outputs and other convex loss functions. More importantly, we show that the same characterization can also be applied to deep ReLU networks with rank-one inputs, where we prove that strong duality still holds and optimal layer weight matrices are rank-one for scalar output networks. As a corollary, we prove that norm regularized deep ReLU networks yield spline interpolation for one-dimensional datasets which was previously known only for two-layer networks. We then verify our theoretical results via several numerical experiments.

arXiv (Cornell University), 2020

Current research in convolutional neural networks (CNN) focuses mainly on changing the architecture of the networks, optimizing the hyper-parameters and improving the gradient descent. However, most work use only 3 standard families of operations inside the CNN, the convolution, the activation function, and the pooling. In this work, we propose a new family of operations based on the Green's function of the Laplacian, which allows the network to solve the Laplacian, to integrate any vector field and to regularize the field by forcing it to be conservative. Hence, the Green's function (GF) is the first operation that regularizes the 2D or 3D feature space by forcing it to be conservative and physically interpretable, instead of regularizing the norm of the weights. Our results show that such regularization allows the network to learn faster, to have smoother training curves and to better generalize, without any additional parameter. The current manuscript presents early results, more work is required to benchmark the proposed method.

2018

In the past few years, deep neural networks have often been claimed to provide greater representational power than shallow networks. In this work, we propose a wide, shallow, and strictly sequential network architecture without any residual connections. When trained with cyclical learning rate schedules, this simple network achieves a classification accuracy on CIFAR100 competitive to a 10 times deeper residual network, while it can be trained 4 times faster. This provides evidence that neither depth nor residual connections are crucial for deep learning. Instead, residual connections just seem to facilitate training using plain SGD by avoiding bad local minima. We believe that our work can hence point the research community to the actual bottleneck of contemporary deep learning: the optimization algorithms.

2022

We study the excess capacity of deep networks in the context of supervised classification. That is, given a capacity measure of the underlying hypothesis class – in our case, Rademacher complexity – how much can we (a-priori) constrain this class while maintaining an empirical error comparable to the unconstrained setting. To assess excess capacity in modern architectures, we first extend an existing generalization bound to accommodate function composition and addition, as well as the specific structure of convolutions. This then facilitates studying residual networks through the lens of the accompanying capacity measure. The key quantities driving this measure are the Lipschitz constants of the layers and the (2,1) group norm distance to the initializations of the convolution weights. We show that these quantities (1) can be kept surprisingly small and, (2) since excess capacity unexpectedly increases with task difficulty, this points towards an unnecessarily large capacity of unco...

Applied Sciences

The main goal of any classification or regression task is to obtain a model that will generalize well on new, previously unseen data. Due to the recent rise of deep learning and many state-of-the-art results obtained with deep models, deep learning architectures have become one of the most used model architectures nowadays. To generalize well, a deep model needs to learn the training data well without overfitting. The latter implies a correlation of deep model optimization and regularization with generalization performance. In this work, we explore the effect of the used optimization algorithm and regularization techniques on the final generalization performance of the model with convolutional neural network (CNN) architecture widely used in the field of computer vision. We give a detailed overview of optimization and regularization techniques with a comparative analysis of their performance with three CNNs on the CIFAR-10 and Fashion-MNIST image datasets.

arXiv (Cornell University), 2019

Tight estimation of the Lipschitz constant for deep neural networks (DNNs) is useful in many applications ranging from robustness certification of classifiers to stability analysis of closed-loop systems with reinforcement learning controllers. Existing methods in the literature for estimating the Lipschitz constant suffer from either lack of accuracy or poor scalability. In this paper, we present a convex optimization framework to compute guaranteed upper bounds on the Lipschitz constant of DNNs both accurately and efficiently. Our main idea is to interpret activation functions as gradients of convex potential functions. Hence, they satisfy certain properties that can be described by quadratic constraints. This particular description allows us to pose the Lipschitz constant estimation problem as a semidefinite program (SDP). The resulting SDP can be adapted to increase either the estimation accuracy (by capturing the interaction between activation functions of different layers) or scalability (by decomposition and parallel implementation). We illustrate the utility of our approach with a variety of experiments on randomly generated networks and on classifiers trained on the MNIST and Iris datasets. In particular, we experimentally demonstrate that our Lipschitz bounds are the most accurate compared to those in the literature. We also study the impact of adversarial training methods on the Lipschitz bounds of the resulting classifiers and show that our bounds can be used to efficiently provide robustness guarantees.