Effective Training Strategies for Deep Graph Neural Networks

Graph Neural Networks (GNNs) tend to suffer performance degradation as model depth increases, which is usually attributed in previous works to the oversmoothing problem. However, we find that although oversmoothing is a contributing factor, the main reasons for this phenomenon are training difficulty and overfitting, which we study by experimentally investigating Graph Convolutional Networks (GCNs), a representative GNN architecture. We find that training difficulty is caused by gradient vanishing and can be solved by adding residual connections. More importantly, overfitting is the major obstacle for deep GCNs and cannot be effectively solved by existing regularization techniques. Deep GCNs also suffer training instability, which slows down the training process. To address overfitting and training instability, we propose Node Normalization (NodeNorm), which normalizes each node using its own statistics in model training. The proposed NodeNorm regularizes deep GCNs by discouraging feature-wise correlation of hidden embeddings and increasing model smoothness with respect to input node features, and thus effectively reduces overfitting. Additionally, it stabilizes the training process and hence speeds up the training. Extensive experiments demonstrate that our NodeNorm method generalizes well to other GNN architectures, enabling deep GNNs to compete with and even outperform shallow ones. Code is publicly available.

Boosting Few-Shot Learning With Adaptive Margin Loss

Few-shot learning (FSL) has attracted increasing attention in recent years but remains challenging, due to the intrinsic difficulty in learning to generalize from a few examples. This paper proposes an adaptive margin principle to improve the generalization ability of metric-based meta-learning approaches for few-shot learning problems. Specifically, we first develop a class-relevant additive margin loss, where semantic similarity between each pair of classes is considered to separate samples in the feature embedding space from similar classes. Further, we incorporate the semantic context among all classes in a sampled training task and develop a task-relevant additive margin loss to better distinguish samples from different classes. Our adaptive margin method can be easily extended to a more realistic generalized FSL setting. Extensive experiments demonstrate that the proposed method can boost the performance of current metric-based meta-learning approaches, under both the standard FSL and generalized FSL settings.

RAIN: Robust and Accurate Classification Networks with Randomization and Enhancement

Along with the extensive applications of CNN models for classification, there has been a growing requirement for their robustness against adversarial examples. In recent years, many adversarial defense methods have been introduced, but most of them have to sacrifice classification accuracy on clean samples to achieve better robustness of CNNs. In this paper, we propose a novel framework to improve robustness and meanwhile retain the accuracy of given classification CNN models, termed as RAIN, which consists of two conjugate modules: structured randomization (SRd) and detail generation (DG). Specifically, the SRd module randomly downsamples and shifts the input, which can destroy the structure of adversarial perturbations so as to improve the model robustness. However, such operations also incur accuracy drop inevitably. Through our empirical study, the resultant image of the SRd module suffers loss of high-frequency details that are crucial for model accuracy. To remedy the accuracy drop, RAIN couples a deep super-resolution model as the DG module for recovering rich details in the resultant image. We evaluate RAIN on STL10 and the ImageNet datasets, and experiment results well demonstrate its great robustness against adversarial examples as well as comparable classification accuracy to non-robustified counterparts on clean samples. Our framework is simple, effective and substantially extends the application of adversarial defense techniques to realistic scenarios where clean and adversarial samples are mixed.

PANDA: Prototypical Unsupervised Domain Adaptation

Previous adversarial domain alignment methods for unsupervised domain adaptation (UDA) pursue conditional domain alignment via intermediate pseudo labels. However, these pseudo labels are generated by independent instances without considering the global data structure and tend to be noisy, making them unreliable for adversarial domain adaptation. Compared with pseudo labels, prototypes are more reliable to represent the data structure resistant to the domain shift since they are summarized over all the relevant instances. In this work, we attempt to calibrate the noisy pseudo labels with prototypes. Specifically, we first obtain a reliable prototypical representation for each instance by multiplying the soft instance predictions with the global prototypes. Based on the prototypical representation, we propose a novel Prototypical Adversarial Learning (PAL) scheme and exploit it to align both feature representations and intermediate prototypes across domains. Besides, with the intermediate prototypes as a proxy, we further minimize the intra-class variance in the target domain to adaptively improve the pseudo labels. Integrating the three objectives, we develop an unified framework termed PrototypicAl uNsupervised Domain Adaptation (PANDA) for UDA. Experiments show that PANDA achieves state-of-the-art or competitive results on multiple UDA benchmarks including both object recognition and semantic segmentation tasks.

Strip Pooling: Rethinking Spatial Pooling for Scene Parsing

Spatial pooling has been proven highly effective in capturing long-range contextual information for pixel-wise prediction tasks, such as scene parsing. In this paper, beyond conventional spatial pooling that usually has a regular shape of NxN, we rethink the formulation of spatial pooling by introducing a new pooling strategy, called strip pooling, which considers a long but narrow kernel, i.e., 1xN or Nx1. Based on strip pooling, we further investigate spatial pooling architecture design by 1) introducing a new strip pooling module that enables backbone networks to efficiently model long-range dependencies, 2) presenting a novel building block with diverse spatial pooling as a core, and 3) systematically comparing the performance of the proposed strip pooling and conventional spatial pooling techniques. Both novel pooling-based designs are lightweight and can serve as an efficient plug-and-play module in existing scene parsing networks. Extensive experiments on popular benchmarks (e.g., ADE20K and Cityscapes) demonstrate that our simple approach establishes new state-of-the-art results. Code is made available at https://github.com/Andrew-Qibin/SPNet.

A Balanced and Uncertainty-aware Approach for Partial Domain Adaptation

This work addresses the unsupervised domain adaptation problem, especially for the partial scenario where the class labels in the target domain are only a subset of those in the source domain. Such a partial transfer setting sounds realistic but challenging while existing methods always suffer from two key problems, i.e., negative transfer and uncertainty propagation. In this paper, we build on domain adversarial learning and propose a novel domain adaptation method BA$^3$US with two new techniques termed Balanced Adversarial Alignment (BAA) and Adaptive Uncertainty Suppression (AUS), respectively. On one hand, negative transfer results in that target samples are misclassified to the classes only present in the source domain. To address this issue, BAA aims to pursue the balance between label distributions across domains in a quite simple manner. Specifically, it randomly leverages a few source samples to augment the smaller target domain during domain alignment so that classes in different domains are symmetric. On the other hand, a source sample is denoted as uncertain if there is an incorrect class that has a relatively high prediction score. Such uncertainty is easily propagated to the unlabeled target data around it during alignment, which severely deteriorates the adaptation performance. Thus, AUS emphasizes uncertain samples and exploits an adaptive weighted complement entropy objective to expect that incorrect classes have the uniform and low prediction scores. Experimental results on multiple benchmarks demonstrate that BA$^3$US surpasses state-of-the-arts for partial domain adaptation tasks.

Cross-layer Feature Pyramid Network for Salient Object Detection

Feature pyramid network (FPN) based models, which fuse the semantics and salient details in a progressive manner, have been proven highly effective in salient object detection. However, it is observed that these models often generate saliency maps with incomplete object structures or unclear object boundaries, due to the \emph{indirect} information propagation among distant layers that makes such fusion structure less effective. In this work, we propose a novel Cross-layer Feature Pyramid Network (CFPN), in which direct cross-layer communication is enabled to improve the progressive fusion in salient object detection. Specifically, the proposed network first aggregates multi-scale features from different layers into feature maps that have access to both the high- and low-level information. Then, it distributes the aggregated features to all the involved layers to gain access to richer context. In this way, the distributed features per layer own both semantics and salient details from all other layers simultaneously, and suffer reduced loss of important information. Extensive experimental results over six widely used salient object detection benchmarks and with three popular backbones clearly demonstrate that CFPN can accurately locate fairly complete salient regions and effectively segment the object boundaries.

Do We Really Need to Access the Source Data? Source Hypothesis Transfer for Unsupervised Domain Adaptation

Unsupervised domain adaptation (UDA) aims to leverage the knowledge learned from a labeled source dataset to solve similar tasks in a new unlabeled domain. Prior UDA methods typically require to access the source data when learning to adapt the model, making them risky and inefficient for decentralized private data. In this work we tackle a novel setting where only a trained source model is available and investigate how we can effectively utilize such a model without source data to solve UDA problems. To this end, we propose a simple yet generic representation learning framework, named \emph{Source HypOthesis Transfer} (SHOT). Specifically, SHOT freezes the classifier module (hypothesis) of the source model and learns the target-specific feature extraction module by exploiting both information maximization and self-supervised pseudo-labeling to implicitly align representations from the target domains to the source hypothesis. In this way, the learned target model can directly predict the labels of target data. We further investigate several techniques to refine the network architecture to parameterize the source model for better transfer performance. To verify its versatility, we evaluate SHOT in a variety of adaptation cases including closed-set, partial-set, and open-set domain adaptation. Experiments indicate that SHOT yields state-of-the-art results among multiple domain adaptation benchmarks.

MetaSelector: Meta-Learning for Recommendation with User-Level Adaptive Model Selection

Recommender systems often face heterogeneous datasets containing highly personalized historical data of users, where no single model could give the best recommendation for every user. We observe this ubiquitous phenomenon on both public and private datasets and address the model selection problem in pursuit of optimizing the quality of recommendation for each user. We propose a meta-learning framework to facilitate user-level adaptive model selection in recommender systems. In this framework, a collection of recommenders is trained with data from all users, on top of which a model selector is trained via meta-learning to select the best single model for each user with the user-specific historical data. We conduct extensive experiments on two public datasets and a real-world production dataset, demonstrating that our proposed framework achieves improvements over single model baselines and sample-level model selector in terms of AUC and LogLoss. In particular, the improvements may lead to huge profit gain when deployed in online recommender systems.