Wenlian Lu

Yi Yu, Wenlian Lu, Boyu Chen

We propose theoretical analyses of a modified natural gradient descent method in the neural network function space based on the eigendecompositions of neural tangent kernel and Fisher information matrix. We firstly present analytical expression for the function learned by this modified natural gradient under the assumptions of Gaussian distribution and infinite width limit. Thus, we explicitly derive the generalization error of the learned neural network function using theoretical methods from eigendecomposition and statistics theory. By decomposing of the total generalization error attributed to different eigenspace of the kernel in function space, we propose a criterion for balancing the errors stemming from training set and the distribution discrepancy between the training set and the true data. Through this approach, we establish that modifying the training direction of the neural network in function space leads to a reduction in the total generalization error. Furthermore, We demonstrate that this theoretical framework is capable to explain many existing results of generalization enhancing methods. These theoretical results are also illustrated by numerical examples on synthetic data.

Junya Chen, Zhe Gan, Xuan Li et al.

InfoNCE-based contrastive representation learners, such as SimCLR, have been tremendously successful in recent years. However, these contrastive schemes are notoriously resource demanding, as their effectiveness breaks down with small-batch training (i.e., the log-K curse, whereas K is the batch-size). In this work, we reveal mathematically why contrastive learners fail in the small-batch-size regime, and present a novel simple, non-trivial contrastive objective named FlatNCE, which fixes this issue. Unlike InfoNCE, our FlatNCE no longer explicitly appeals to a discriminative classification goal for contrastive learning. Theoretically, we show FlatNCE is the mathematical dual formulation of InfoNCE, thus bridging the classical literature on energy modeling; and empirically, we demonstrate that, with minimal modification of code, FlatNCE enables immediate performance boost independent of the subject-matter engineering efforts. The significance of this work is furthered by the powerful generalization of contrastive learning techniques, and the introduction of new tools to monitor and diagnose contrastive training. We substantiate our claims with empirical evidence on CIFAR10, ImageNet, and other datasets, where FlatNCE consistently outperforms InfoNCE.

Boyu Chen, Wenlian Lu, Ernest Fokoue

Meta-learning is a promising method to achieve efficient training method towards deep neural net and has been attracting increases interests in recent years. But most of the current methods are still not capable to train complex neuron net model with long-time training process. In this paper, a novel second-order meta-optimizer, named Meta-learning with Hessian-Free(MLHF) approach, is proposed based on the Hessian-Free approach. Two recurrent neural networks are established to generate the damping and the precondition matrix of this Hessian-Free framework. A series of techniques to meta-train the MLHF towards stable and reinforce the meta-training of this optimizer, including the gradient calculation of $H$. Numerical experiments on deep convolution neural nets, including CUDA-convnet and ResNet18(v2), with datasets of CIFAR10 and ILSVRC2012, indicate that the MLHF shows good and continuous training performance during the whole long-time training process, i.e., both the rapid-decreasing early stage and the steadily-deceasing later stage, and so is a promising meta-learning framework towards elevating the training efficiency in real-world deep neural nets.

Shouhuai Xu, Wenlian Lu, Li Xu

Understanding the dynamics of computer virus (malware, worm) in cyberspace is an important problem that has attracted a fair amount of attention. Early investigations for this purpose adapted biological epidemic models, and thus inherited the so-called homogeneity assumption that each node is equally connected to others. Later studies relaxed this often-unrealistic homogeneity assumption, but still focused on certain power-law networks. Recently, researchers investigated epidemic models in {\em arbitrary} networks (i.e., no restrictions on network topology). However, all these models only capture {\em push-based} infection, namely that an infectious node always actively attempts to infect its neighboring nodes. Very recently, the concept of {\em pull-based} infection was introduced but was not treated rigorously. Along this line of research, the present paper investigates push- and pull-based epidemic spreading dynamics in arbitrary networks, using a Non-linear Dynamical Systems approach. The paper advances the state of the art as follows: (1) It presents a more general and powerful sufficient condition (also known as epidemic threshold in the literature) under which the spreading will become stable. (2) It gives both upper and lower bounds on the global mean infection rate, regardless of the stability of the spreading. (3) It offers insights into, among other things, the estimation of the global mean infection rate through localized monitoring of a small {\em constant} number of nodes, {\em without} knowing the values of the parameters.