Abstract:This paper demonstrates that in classification problems, fully connected neural networks (FCNs) and residual neural networks (ResNets) cannot be approximated by kernel logistic regression based on the Neural Tangent Kernel (NTK) under overtraining (i.e., when training time approaches infinity). Specifically, when using the cross-entropy loss, regardless of how large the network width is (as long as it is finite), the empirical NTK diverges from the NTK on the training samples as training time increases. To establish this result, we first demonstrate the strictly positive definiteness of the NTKs for multi-layer FCNs and ResNets. Then, we prove that during training, % with the cross-entropy loss, the neural network parameters diverge if the smallest eigenvalue of the empirical NTK matrix (Gram matrix) with respect to training samples is bounded below by a positive constant. This behavior contrasts sharply with the lazy training regime commonly observed in regression problems. Consequently, using a proof by contradiction, we show that the empirical NTK does not uniformly converge to the NTK across all times on the training samples as the network width increases. We validate our theoretical results through experiments on both synthetic data and the MNIST classification task. This finding implies that NTK theory is not applicable in this context, with significant theoretical implications for understanding neural networks in classification problems.
Abstract:Deep Residual Neural Networks (ResNets) have demonstrated remarkable success across a wide range of real-world applications. In this paper, we identify a suitable scaling factor (denoted by $\alpha$) on the residual branch of deep wide ResNets to achieve good generalization ability. We show that if $\alpha$ is a constant, the class of functions induced by Residual Neural Tangent Kernel (RNTK) is asymptotically not learnable, as the depth goes to infinity. We also highlight a surprising phenomenon: even if we allow $\alpha$ to decrease with increasing depth $L$, the degeneration phenomenon may still occur. However, when $\alpha$ decreases rapidly with $L$, the kernel regression with deep RNTK with early stopping can achieve the minimax rate provided that the target regression function falls in the reproducing kernel Hilbert space associated with the infinite-depth RNTK. Our simulation studies on synthetic data and real classification tasks such as MNIST, CIFAR10 and CIFAR100 support our theoretical criteria for choosing $\alpha$.
Abstract:In this paper, we study the generalization ability of the wide residual network on $\mathbb{S}^{d-1}$ with the ReLU activation function. We first show that as the width $m\rightarrow\infty$, the residual network kernel (RNK) uniformly converges to the residual neural tangent kernel (RNTK). This uniform convergence further guarantees that the generalization error of the residual network converges to that of the kernel regression with respect to the RNTK. As direct corollaries, we then show $i)$ the wide residual network with the early stopping strategy can achieve the minimax rate provided that the target regression function falls in the reproducing kernel Hilbert space (RKHS) associated with the RNTK; $ii)$ the wide residual network can not generalize well if it is trained till overfitting the data. We finally illustrate some experiments to reconcile the contradiction between our theoretical result and the widely observed ``benign overfitting phenomenon''