In recent years, graph neural networks (GNNs) have emerged as a successful tool in a variety of graph-related applications. However, the performance of GNNs can be deteriorated when noisy connections occur in the original graph structures; besides, the dependence on explicit structures prevents GNNs from being applied to general unstructured scenarios. To address these issues, recently emerged deep graph structure learning (GSL) methods propose to jointly optimize the graph structure along with GNN under the supervision of a node classification task. Nonetheless, these methods focus on a supervised learning scenario, which leads to several problems, i.e., the reliance on labels, the bias of edge distribution, and the limitation on application tasks. In this paper, we propose a more practical GSL paradigm, unsupervised graph structure learning, where the learned graph topology is optimized by data itself without any external guidance (i.e., labels). To solve the unsupervised GSL problem, we propose a novel StrUcture Bootstrapping contrastive LearnIng fraMEwork (SUBLIME for abbreviation) with the aid of self-supervised contrastive learning. Specifically, we generate a learning target from the original data as an "anchor graph", and use a contrastive loss to maximize the agreement between the anchor graph and the learned graph. To provide persistent guidance, we design a novel bootstrapping mechanism that upgrades the anchor graph with learned structures during model learning. We also design a series of graph learners and post-processing schemes to model the structures to learn. Extensive experiments on eight benchmark datasets demonstrate the significant effectiveness of our proposed SUBLIME and high quality of the optimized graphs.
Recently, unsupervised person re-identification (Re-ID) has received increasing research attention due to its potential for label-free applications. A promising way to address unsupervised Re-ID is clustering-based, which generates pseudo labels by clustering and uses the pseudo labels to train a Re-ID model iteratively. However, most clustering-based methods take each cluster as a pseudo identity class, neglecting the intra-cluster variance mainly caused by the change of cameras. To address this issue, we propose to split each single cluster into multiple proxies according to camera views. The camera-aware proxies explicitly capture local structures within clusters, by which the intra-ID variance and inter-ID similarity can be better tackled. Assisted with the camera-aware proxies, we design two proxy-level contrastive learning losses that are, respectively, based on offline and online association results. The offline association directly associates proxies according to the clustering and splitting results, while the online strategy dynamically associates proxies in terms of up-to-date features to reduce the noise caused by the delayed update of pseudo labels. The combination of two losses enables us to train a desirable Re-ID model. Extensive experiments on three person Re-ID datasets and one vehicle Re-ID dataset show that our proposed approach demonstrates competitive performance with state-of-the-art methods. Code will be available at: https://github.com/Terminator8758/O2CAP.
Weakly supervised semantic segmentation (WSSS) has gained significant popularity since it relies only on weak labels such as image level annotations rather than pixel level annotations required by supervised semantic segmentation (SSS) methods. Despite drastically reduced annotation costs, typical feature representations learned from WSSS are only representative of some salient parts of objects and less reliable compared to SSS due to the weak guidance during training. In this paper, we propose a novel Multi-Strategy Contrastive Learning (MuSCLe) framework to obtain enhanced feature representations and improve WSSS performance by exploiting similarity and dissimilarity of contrastive sample pairs at image, region, pixel and object boundary levels. Extensive experiments demonstrate the effectiveness of our method and show that MuSCLe outperforms the current state-of-the-art on the widely used PASCAL VOC 2012 dataset.
Self-supervised entity alignment (EA) aims to link equivalent entities across different knowledge graphs (KGs) without seed alignments. The current SOTA self-supervised EA method draws inspiration from contrastive learning, originally designed in computer vision based on instance discrimination and contrastive loss, and suffers from two shortcomings. Firstly, it puts unidirectional emphasis on pushing sampled negative entities far away rather than pulling positively aligned pairs close, as is done in the well-established supervised EA. Secondly, KGs contain rich side information (e.g., entity description), and how to effectively leverage those information has not been adequately investigated in self-supervised EA. In this paper, we propose an interactive contrastive learning model for self-supervised EA. The model encodes not only structures and semantics of entities (including entity name, entity description, and entity neighborhood), but also conducts cross-KG contrastive learning by building pseudo-aligned entity pairs. Experimental results show that our approach outperforms previous best self-supervised results by a large margin (over 9% average improvement) and performs on par with previous SOTA supervised counterparts, demonstrating the effectiveness of the interactive contrastive learning for self-supervised EA.
Recently, self-supervised vision transformers have attracted unprecedented attention for their impressive representation learning ability. However, the dominant method, contrastive learning, mainly relies on an instance discrimination pretext task, which learns a global understanding of the image. This paper incorporates local feature learning into self-supervised vision transformers via Reconstructive Pre-training (RePre). Our RePre extends contrastive frameworks by adding a branch for reconstructing raw image pixels in parallel with the existing contrastive objective. RePre is equipped with a lightweight convolution-based decoder that fuses the multi-hierarchy features from the transformer encoder. The multi-hierarchy features provide rich supervisions from low to high semantic information, which are crucial for our RePre. Our RePre brings decent improvements on various contrastive frameworks with different vision transformer architectures. Transfer performance in downstream tasks outperforms supervised pre-training and state-of-the-art (SOTA) self-supervised counterparts.
Graph contrastive learning attracts/disperses node representations for similar/dissimilar node pairs under some notion of similarity. It may be combined with a low-dimensional embedding of nodes to preserve intrinsic and structural properties of a graph. In this paper, we extend the celebrated Laplacian Eigenmaps with contrastive learning, and call them COntrastive Laplacian EigenmapS (COLES). Starting from a GAN-inspired contrastive formulation, we show that the Jensen-Shannon divergence underlying many contrastive graph embedding models fails under disjoint positive and negative distributions, which may naturally emerge during sampling in the contrastive setting. In contrast, we demonstrate analytically that COLES essentially minimizes a surrogate of Wasserstein distance, which is known to cope well under disjoint distributions. Moreover, we show that the loss of COLES belongs to the family of so-called block-contrastive losses, previously shown to be superior compared to pair-wise losses typically used by contrastive methods. We show on popular benchmarks/backbones that COLES offers favourable accuracy/scalability compared to DeepWalk, GCN, Graph2Gauss, DGI and GRACE baselines.
Recent advances on prompt-tuning cast few-shot classification tasks as a masked language modeling problem. By wrapping input into a template and using a verbalizer which constructs a mapping between label space and label word space, prompt-tuning can achieve excellent results in zero-shot and few-shot scenarios. However, typical prompt-tuning needs a manually designed verbalizer which requires domain expertise and human efforts. And the insufficient label space may introduce considerable bias into the results. In this paper, we focus on eliciting knowledge from pretrained language models and propose a prototypical prompt verbalizer for prompt-tuning. Labels are represented by prototypical embeddings in the feature space rather than by discrete words. The distances between the embedding at the masked position of input and prototypical embeddings are used as classification criterion. For zero-shot settings, knowledge is elicited from pretrained language models by a manually designed template to form initial prototypical embeddings. For few-shot settings, models are tuned to learn meaningful and interpretable prototypical embeddings. Our method optimizes models by contrastive learning. Extensive experimental results on several many-class text classification datasets with low-resource settings demonstrate the effectiveness of our approach compared with other verbalizer construction methods. Our implementation is available at https://github.com/Ydongd/prototypical-prompt-verbalizer.
Vision-language (V+L) pretraining models have achieved great success in supporting multimedia applications by understanding the alignments between images and text. While existing vision-language pretraining models primarily focus on understanding objects in images or entities in text, they often ignore the alignment at the level of events and their argument structures. In this work, we propose a contrastive learning framework to enforce vision-language pretraining models to comprehend events and associated argument (participant) roles. To achieve this, we take advantage of text information extraction technologies to obtain event structural knowledge, and utilize multiple prompt functions to contrast difficult negative descriptions by manipulating event structures. We also design an event graph alignment loss based on optimal transport to capture event argument structures. In addition, we collect a large event-rich dataset (106,875 images) for pretraining, which provides a more challenging image retrieval benchmark to assess the understanding of complicated lengthy sentences. Experiments show that our zero-shot CLIP-Event outperforms the state-of-the-art supervised model in argument extraction on Multimedia Event Extraction, achieving more than 5% absolute F-score gain in event extraction, as well as significant improvements on a variety of downstream tasks under zero-shot settings.
Semantic networks, such as the knowledge graph, can represent the knowledge leveraging the graph structure. Although the knowledge graph shows promising values in natural language processing, it suffers from incompleteness. This paper focuses on knowledge graph completion by predicting linkage between entities, which is a fundamental yet critical task. Semantic matching is a potential solution as it can deal with unseen entities, which the translational distance based methods struggle with. However, to achieve competitive performance as translational distance based methods, semantic matching based methods require large-scale datasets for the training purpose, which are typically unavailable in practical settings. Therefore, we employ the language model and introduce a novel knowledge graph architecture named LP-BERT, which contains two main stages: multi-task pre-training and knowledge graph fine-tuning. In the pre-training phase, three tasks are taken to drive the model to learn the relationship from triples by predicting either entities or relations. While in the fine-tuning phase, inspired by contrastive learning, we design a triple-style negative sampling in a batch, which greatly increases the proportion of negative sampling while keeping the training time almost unchanged. Furthermore, we propose a new data augmentation method utilizing the inverse relationship of triples to improve the performance and robustness of the model. To demonstrate the effectiveness of our method, we conduct extensive experiments on three widely-used datasets, WN18RR, FB15k-237, and UMLS. The experimental results demonstrate the superiority of our methods, and our approach achieves state-of-the-art results on WN18RR and FB15k-237 datasets. Significantly, Hits@10 indicator is improved by 5% from previous state-of-the-art result on the WN18RR dataset while reaching 100% on the UMLS dataset.
Manually annotating complex scene point cloud datasets is both costly and error-prone. To reduce the reliance on labeled data, a new model called SnapshotNet is proposed as a self-supervised feature learning approach, which directly works on the unlabeled point cloud data of a complex 3D scene. The SnapshotNet pipeline includes three stages. In the snapshot capturing stage, snapshots, which are defined as local collections of points, are sampled from the point cloud scene. A snapshot could be a view of a local 3D scan directly captured from the real scene, or a virtual view of such from a large 3D point cloud dataset. Snapshots could also be sampled at different sampling rates or fields of view (FOVs), thus multi-FOV snapshots, to capture scale information from the scene. In the feature learning stage, a new pre-text task called multi-FOV contrasting is proposed to recognize whether two snapshots are from the same object or not, within the same FOV or across different FOVs. Snapshots go through two self-supervised learning steps: the contrastive learning step with both part and scale contrasting, followed by a snapshot clustering step to extract higher level semantic features. Then a weakly-supervised segmentation stage is implemented by first training a standard SVM classifier on the learned features with a small fraction of labeled snapshots. The trained SVM is used to predict labels for input snapshots and predicted labels are converted into point-wise label assignments for semantic segmentation of the entire scene using a voting procedure. The experiments are conducted on the Semantic3D dataset and the results have shown that the proposed method is capable of learning effective features from snapshots of complex scene data without any labels. Moreover, the proposed method has shown advantages when comparing to the SOA method on weakly-supervised point cloud semantic segmentation.
We propose PromptBERT, a novel contrastive learning method for learning better sentence representation. We firstly analyze the drawback of current sentence embedding from original BERT and find that it is mainly due to the static token embedding bias and ineffective BERT layers. Then we propose the first prompt-based sentence embeddings method and discuss two prompt representing methods and three prompt searching methods to make BERT achieve better sentence embeddings. Moreover, we propose a novel unsupervised training objective by the technology of template denoising, which substantially shortens the performance gap between the supervised and unsupervised settings. Extensive experiments show the effectiveness of our method. Compared to SimCSE, PromptBert achieves 2.29 and 2.58 points of improvement based on BERT and RoBERTa in the unsupervised setting.
Contrastive learning relies on an assumption that positive pairs contain related views, e.g., patches of an image or co-occurring multimodal signals of a video, that share certain underlying information about an instance. But what if this assumption is violated? The literature suggests that contrastive learning produces suboptimal representations in the presence of noisy views, e.g., false positive pairs with no apparent shared information. In this work, we propose a new contrastive loss function that is robust against noisy views. We provide rigorous theoretical justifications by showing connections to robust symmetric losses for noisy binary classification and by establishing a new contrastive bound for mutual information maximization based on the Wasserstein distance measure. The proposed loss is completely modality-agnostic and a simple drop-in replacement for the InfoNCE loss, which makes it easy to apply to existing contrastive frameworks. We show that our approach provides consistent improvements over the state-of-the-art on image, video, and graph contrastive learning benchmarks that exhibit a variety of real-world noise patterns.
The unsupervised domain adaptive person re-identification (re-ID) task has been a challenge because, unlike the general domain adaptive tasks, there is no overlap between the classes of source and target domain data in the person re-ID, which leads to a significant domain gap. State-of-the-art unsupervised re-ID methods train the neural networks using a memory-based contrastive loss. However, performing contrastive learning by treating each unlabeled instance as a class will lead to the problem of class collision, and the updating intensity is inconsistent due to the difference in the number of instances of different categories when updating in the memory bank. To address such problems, we propose Prototype Dictionary Learning for person re-ID which is able to utilize both source domain data and target domain data by one training stage while avoiding the problem of class collision and the problem of updating intensity inconsistency by cluster-level prototype dictionary learning. In order to reduce the interference of domain gap on the model, we propose a local-enhance module to improve the domain adaptation of the model without increasing the number of model parameters. Our experiments on two large datasets demonstrate the effectiveness of the prototype dictionary learning. 71.5\% mAP is achieved in the Market-to-Duke task, which is a 2.3\% improvement compared to the state-of-the-art unsupervised domain adaptive re-ID methods. It achieves 83.9\% mAP in the Duke-to-Market task, which improves by 4.4\% compared to the state-of-the-art unsupervised adaptive re-ID methods.
Motion, as the most distinct phenomenon in a video to involve the changes over time, has been unique and critical to the development of video representation learning. In this paper, we ask the question: how important is the motion particularly for self-supervised video representation learning. To this end, we compose a duet of exploiting the motion for data augmentation and feature learning in the regime of contrastive learning. Specifically, we present a Motion-focused Contrastive Learning (MCL) method that regards such duet as the foundation. On one hand, MCL capitalizes on optical flow of each frame in a video to temporally and spatially sample the tubelets (i.e., sequences of associated frame patches across time) as data augmentations. On the other hand, MCL further aligns gradient maps of the convolutional layers to optical flow maps from spatial, temporal and spatio-temporal perspectives, in order to ground motion information in feature learning. Extensive experiments conducted on R(2+1)D backbone demonstrate the effectiveness of our MCL. On UCF101, the linear classifier trained on the representations learnt by MCL achieves 81.91% top-1 accuracy, outperforming ImageNet supervised pre-training by 6.78%. On Kinetics-400, MCL achieves 66.62% top-1 accuracy under the linear protocol. Code is available at https://github.com/YihengZhang-CV/MCL-Motion-Focused-Contrastive-Learning.
In this study, we propose a feature extraction framework based on contrastive learning with adaptive positive and negative samples (CL-FEFA) that is suitable for unsupervised, supervised, and semi-supervised single-view feature extraction. CL-FEFA constructs adaptively the positive and negative samples from the results of feature extraction, which makes it more appropriate and accurate. Thereafter, the discriminative features are re extracted to according to InfoNCE loss based on previous positive and negative samples, which will make the intra-class samples more compact and the inter-class samples more dispersed. At the same time, using the potential structure information of subspace samples to dynamically construct positive and negative samples can make our framework more robust to noisy data. Furthermore, CL-FEFA considers the mutual information between positive samples, that is, similar samples in potential structures, which provides theoretical support for its advantages in feature extraction. The final numerical experiments prove that the proposed framework has a strong advantage over the traditional feature extraction methods and contrastive learning methods.
Recent works explore learning graph representations in a self-supervised manner. In graph contrastive learning, benchmark methods apply various graph augmentation approaches. However, most of the augmentation methods are non-learnable, which causes the issue of generating unbeneficial augmented graphs. Such augmentation may degenerate the representation ability of graph contrastive learning methods. Therefore, we motivate our method to generate augmented graph by a learnable graph augmenter, called MEta Graph Augmentation (MEGA). We then clarify that a "good" graph augmentation must have uniformity at the instance-level and informativeness at the feature-level. To this end, we propose a novel approach to learning a graph augmenter that can generate an augmentation with uniformity and informativeness. The objective of the graph augmenter is to promote our feature extraction network to learn a more discriminative feature representation, which motivates us to propose a meta-learning paradigm. Empirically, the experiments across multiple benchmark datasets demonstrate that MEGA outperforms the state-of-the-art methods in graph self-supervised learning tasks. Further experimental studies prove the effectiveness of different terms of MEGA.
Glaucoma is one of the ophthalmic diseases that may cause blindness, for which early detection and treatment are very important. Fundus images and optical coherence tomography (OCT) images are both widely-used modalities in diagnosing glaucoma. However, existing glaucoma grading approaches mainly utilize a single modality, ignoring the complementary information between fundus and OCT. In this paper, we propose an efficient multi-modality supervised contrastive learning framework, named COROLLA, for glaucoma grading. Through layer segmentation as well as thickness calculation and projection, retinal thickness maps are extracted from the original OCT volumes and used as a replacing modality, resulting in more efficient calculations with less memory usage. Given the high structure and distribution similarities across medical image samples, we employ supervised contrastive learning to increase our models' discriminative power with better convergence. Moreover, feature-level fusion of paired fundus image and thickness map is conducted for enhanced diagnosis accuracy. On the GAMMA dataset, our COROLLA framework achieves overwhelming glaucoma grading performance compared to state-of-the-art methods.
Non-Local Attention (NLA) brings significant improvement for Single Image Super-Resolution (SISR) by leveraging intrinsic feature correlation in natural images. However, NLA gives noisy information large weights and consumes quadratic computation resources with respect to the input size, limiting its performance and application. In this paper, we propose a novel Efficient Non-Local Contrastive Attention (ENLCA) to perform long-range visual modeling and leverage more relevant non-local features. Specifically, ENLCA consists of two parts, Efficient Non-Local Attention (ENLA) and Sparse Aggregation. ENLA adopts the kernel method to approximate exponential function and obtains linear computation complexity. For Sparse Aggregation, we multiply inputs by an amplification factor to focus on informative features, yet the variance of approximation increases exponentially. Therefore, contrastive learning is applied to further separate relevant and irrelevant features. To demonstrate the effectiveness of ENLCA, we build an architecture called Efficient Non-Local Contrastive Network (ENLCN) by adding a few of our modules in a simple backbone. Extensive experimental results show that ENLCN reaches superior performance over state-of-the-art approaches on both quantitative and qualitative evaluations.
In this work, we aim to consider the application of contrastive learning in the scenario of the recommendation system adequately, making it more suitable for recommendation task. We propose a learning paradigm called supervised contrastive learning(SCL) to support the graph convolutional neural network. Specifically, we will calculate the similarity between different nodes in user side and item side respectively during data preprocessing, and then when applying contrastive learning, not only will the augmented views be regarded as the positive samples, but also a certain number of similar samples will be regarded as the positive samples, which is different with SimCLR that treats other samples in a batch as negative samples. We apply SCL on the most advanced LightGCN. In addition, in order to consider the uncertainty of node interaction, we also propose a new data augment method called node replication. Empirical research and ablation study on Gowalla, Yelp2018, Amazon-Book datasets prove the effectiveness of SCL and node replication, which improve the accuracy of recommendations and robustness to interactive noise.
Conversational recommender systems (CRS) aim to recommend suitable items to users through natural language conversations. For developing effective CRSs, a major technical issue is how to accurately infer user preference from very limited conversation context. To address issue, a promising solution is to incorporate external data for enriching the context information. However, prior studies mainly focus on designing fusion models tailored for some specific type of external data, which is not general to model and utilize multi-type external data. To effectively leverage multi-type external data, we propose a novel coarse-to-fine contrastive learning framework to improve data semantic fusion for CRS. In our approach, we first extract and represent multi-grained semantic units from different data signals, and then align the associated multi-type semantic units in a coarse-to-fine way. To implement this framework, we design both coarse-grained and fine-grained procedures for modeling user preference, where the former focuses on more general, coarse-grained semantic fusion and the latter focuses on more specific, fine-grained semantic fusion. Such an approach can be extended to incorporate more kinds of external data. Extensive experiments on two public CRS datasets have demonstrated the effectiveness of our approach in both recommendation and conversation tasks.