Self-supervised learning with masked autoencoders has recently gained popularity for its ability to produce effective image or textual representations, which can be applied to various downstream tasks without retraining. However, we observe that the current masked autoencoder models lack good generalization ability on graph data. To tackle this issue, we propose a novel graph masked autoencoder framework called GiGaMAE. Different from existing masked autoencoders that learn node presentations by explicitly reconstructing the original graph components (e.g., features or edges), in this paper, we propose to collaboratively reconstruct informative and integrated latent embeddings. By considering embeddings encompassing graph topology and attribute information as reconstruction targets, our model could capture more generalized and comprehensive knowledge. Furthermore, we introduce a mutual information based reconstruction loss that enables the effective reconstruction of multiple targets. This learning objective allows us to differentiate between the exclusive knowledge learned from a single target and common knowledge shared by multiple targets. We evaluate our method on three downstream tasks with seven datasets as benchmarks. Extensive experiments demonstrate the superiority of GiGaMAE against state-of-the-art baselines. We hope our results will shed light on the design of foundation models on graph-structured data. Our code is available at: https://github.com/sycny/GiGaMAE.
With the success of self-supervised learning, multimodal foundation models have rapidly adapted a wide range of downstream tasks driven by vision and language (VL) pretraining. State-of-the-art methods achieve impressive performance by pre-training on large-scale datasets. However, bridging the semantic gap between the two modalities remains a nonnegligible challenge for VL tasks. In this work, we propose an efficient computation framework for multimodal alignment by introducing a novel visual semantic module to further improve the performance of the VL tasks. Specifically, we propose a flexible model, namely Artificial-Spiking Hierarchical Networks (ASH-Nets), which combines the complementary advantages of Artificial neural networks (ANNs) and Spiking neural networks (SNNs) to enrich visual semantic representations. In particular, a visual concrete encoder and a semantic abstract encoder are constructed to learn continuous and discrete latent variables to enhance the flexibility of semantic encoding. Considering the spatio-temporal properties of SNNs modeling, we introduce a contrastive learning method to optimize the inputs of similar samples. This can improve the computational efficiency of the hierarchical network, while the augmentation of hard samples is beneficial to the learning of visual representations. Furthermore, the Spiking to Text Uni-Alignment Learning (STUA) pre-training method is proposed, which only relies on text features to enhance the encoding ability of abstract semantics. We validate the performance on multiple well-established downstream VL tasks. Experiments show that the proposed ASH-Nets achieve competitive results.
We propose a unified point cloud video self-supervised learning framework for object-centric and scene-centric data. Previous methods commonly conduct representation learning at the clip or frame level and cannot well capture fine-grained semantics. Instead of contrasting the representations of clips or frames, in this paper, we propose a unified self-supervised framework by conducting contrastive learning at the point level. Moreover, we introduce a new pretext task by achieving semantic alignment of superpoints, which further facilitates the representations to capture semantic cues at multiple scales. In addition, due to the high redundancy in the temporal dimension of dynamic point clouds, directly conducting contrastive learning at the point level usually leads to massive undesired negatives and insufficient modeling of positive representations. To remedy this, we propose a selection strategy to retain proper negatives and make use of high-similarity samples from other instances as positive supplements. Extensive experiments show that our method outperforms supervised counterparts on a wide range of downstream tasks and demonstrates the superior transferability of the learned representations.
Recently, the community has made tremendous progress in developing effective methods for point cloud video understanding that learn from massive amounts of labeled data. However, annotating point cloud videos is usually notoriously expensive. Moreover, training via one or only a few traditional tasks (e.g., classification) may be insufficient to learn subtle details of the spatio-temporal structure existing in point cloud videos. In this paper, we propose a Masked Spatio-Temporal Structure Prediction (MaST-Pre) method to capture the structure of point cloud videos without human annotations. MaST-Pre is based on spatio-temporal point-tube masking and consists of two self-supervised learning tasks. First, by reconstructing masked point tubes, our method is able to capture the appearance information of point cloud videos. Second, to learn motion, we propose a temporal cardinality difference prediction task that estimates the change in the number of points within a point tube. In this way, MaST-Pre is forced to model the spatial and temporal structure in point cloud videos. Extensive experiments on MSRAction-3D, NTU-RGBD, NvGesture, and SHREC'17 demonstrate the effectiveness of the proposed method.
This report describes ID R&D team submissions for Track 2 (open) to the VoxCeleb Speaker Recognition Challenge 2023 (VoxSRC-23). Our solution is based on the fusion of deep ResNets and self-supervised learning (SSL) based models trained on a mixture of a VoxCeleb2 dataset and a large version of a VoxTube dataset. The final submission to the Track 2 achieved the first place on the VoxSRC-23 public leaderboard with a minDCF(0.05) of 0.0762 and EER of 1.30%.
Multivariate time-series forecasting is vital in various domains, e.g., economic planning and weather prediction. Deep train-from-scratch models have exhibited effective performance yet require large amounts of data, which limits real-world applicability. Recently, researchers have leveraged the representation learning transferability of pre-trained Large Language Models (LLMs) to handle limited non-linguistic datasets effectively. However, incorporating LLMs with time-series data presents challenges of limited adaptation due to different compositions between time-series and linguistic data, and the inability to process multi-scale temporal information. To tackle these challenges, we propose LLM4TS, a framework for time-series forecasting with pre-trained LLMs. LLM4TS consists of a two-stage fine-tuning strategy: the \textit{time-series alignment} stage to align LLMs with the nuances of time-series data, and the \textit{forecasting fine-tuning} stage for downstream time-series forecasting tasks. Furthermore, our framework features a novel two-level aggregation method that integrates multi-scale temporal data within pre-trained LLMs, enhancing their ability to interpret time-specific information. In experiments across 7 time-series forecasting datasets, LLM4TS is superior to existing state-of-the-art methods compared with trained-from-scratch models in full-shot scenarios, and also achieves an average improvement of 6.84% in MSE in few-shot scenarios. In addition, evaluations compared with different self-supervised learning approaches highlight LLM4TS's effectiveness with representation learning in forecasting tasks.
Detecting lane markings in road scenes poses a challenge due to their intricate nature, which is susceptible to unfavorable conditions. While lane markings have strong shape priors, their visibility is easily compromised by lighting conditions, occlusions by other vehicles or pedestrians, and fading of colors over time. The detection process is further complicated by the presence of several lane shapes and natural variations, necessitating large amounts of data to train a robust lane detection model capable of handling various scenarios. In this paper, we present a novel self-supervised learning method termed Contrastive Learning for Lane Detection via cross-similarity (CLLD) to enhance the resilience of lane detection models in real-world scenarios, particularly when the visibility of lanes is compromised. CLLD introduces a contrastive learning (CL) method that assesses the similarity of local features within the global context of the input image. It uses the surrounding information to predict lane markings. This is achieved by integrating local feature contrastive learning with our proposed cross-similar operation. The local feature CL concentrates on extracting features from small patches, a necessity for accurately localizing lane segments. Meanwhile, cross-similarity captures global features, enabling the detection of obscured lane segments based on their surroundings. We enhance cross-similarity by randomly masking portions of input images in the process of augmentation. Extensive experiments on TuSimple and CuLane benchmarks demonstrate that CLLD outperforms SOTA contrastive learning methods, particularly in visibility-impairing conditions like shadows, while it also delivers comparable results under normal conditions. Compared to supervised learning, CLLD still excels in challenging scenarios such as shadows and crowded scenes, which are common in real-world driving.
Knowledge graphs (KGs) are commonly used as side information to enhance collaborative signals and improve recommendation quality. In the context of knowledge-aware recommendation (KGR), graph neural networks (GNNs) have emerged as promising solutions for modeling factual and semantic information in KGs. However, the long-tail distribution of entities leads to sparsity in supervision signals, which weakens the quality of item representation when utilizing KG enhancement. Additionally, the binary relation representation of KGs simplifies hyper-relational facts, making it challenging to model complex real-world information. Furthermore, the over-smoothing phenomenon results in indistinguishable representations and information loss. To address these challenges, we propose the SDK (Self-Supervised Dynamic Hypergraph Recommendation based on Hyper-Relational Knowledge Graph) framework. This framework establishes a cross-view hypergraph self-supervised learning mechanism for KG enhancement. Specifically, we model hyper-relational facts in KGs to capture interdependencies between entities under complete semantic conditions. With the refined representation, a hypergraph is dynamically constructed to preserve features in the deep vector space, thereby alleviating the over-smoothing problem. Furthermore, we mine external supervision signals from both the global perspective of the hypergraph and the local perspective of collaborative filtering (CF) to guide the model prediction process. Extensive experiments conducted on different datasets demonstrate the superiority of the SDK framework over state-of-the-art models. The results showcase its ability to alleviate the effects of over-smoothing and supervision signal sparsity.
Pre-training datasets, like ImageNet, have become the gold standard in medical image analysis. However, the emergence of self-supervised learning (SSL), which leverages unlabeled data to learn robust features, presents an opportunity to bypass the intensive labeling process. In this study, we explored if SSL for pre-training on non-medical images can be applied to chest radiographs and how it compares to supervised pre-training on non-medical images and on medical images. We utilized a vision transformer and initialized its weights based on (i) SSL pre-training on natural images (DINOv2), (ii) SL pre-training on natural images (ImageNet dataset), and (iii) SL pre-training on chest radiographs from the MIMIC-CXR database. We tested our approach on over 800,000 chest radiographs from six large global datasets, diagnosing more than 20 different imaging findings. Our SSL pre-training on curated images not only outperformed ImageNet-based pre-training (P<0.001 for all datasets) but, in certain cases, also exceeded SL on the MIMIC-CXR dataset. Our findings suggest that selecting the right pre-training strategy, especially with SSL, can be pivotal for improving artificial intelligence (AI)'s diagnostic accuracy in medical imaging. By demonstrating the promise of SSL in chest radiograph analysis, we underline a transformative shift towards more efficient and accurate AI models in medical imaging.
We present a method for learning multiple scene representations given a small labeled set, by exploiting the relationships between such representations in the form of a multi-task hypergraph. We also show how we can use the hypergraph to improve a powerful pretrained VisTransformer model without any additional labeled data. In our hypergraph, each node is an interpretation layer (e.g., depth or segmentation) of the scene. Within each hyperedge, one or several input nodes predict the layer at the output node. Thus, each node could be an input node in some hyperedges and an output node in others. In this way, multiple paths can reach the same node, to form ensembles from which we obtain robust pseudolabels, which allow self-supervised learning in the hypergraph. We test different ensemble models and different types of hyperedges and show superior performance to other multi-task graph models in the field. We also introduce Dronescapes, a large video dataset captured with UAVs in different complex real-world scenes, with multiple representations, suitable for multi-task learning.
Deep learning in medical imaging has the potential to minimize the risk of diagnostic errors, reduce radiologist workload, and accelerate diagnosis. Training such deep learning models requires large and accurate datasets, with annotations for all training samples. However, in the medical imaging domain, annotated datasets for specific tasks are often small due to the high complexity of annotations, limited access, or the rarity of diseases. To address this challenge, deep learning models can be pre-trained on large image datasets without annotations using methods from the field of self-supervised learning. After pre-training, small annotated datasets are sufficient to fine-tune the models for a specific task. The most popular self-supervised pre-training approaches in medical imaging are based on contrastive learning. However, recent studies in natural image processing indicate a strong potential for masked autoencoder approaches. Our work compares state-of-the-art contrastive learning methods with the recently introduced masked autoencoder approach "SparK" for convolutional neural networks (CNNs) on medical images. Therefore we pre-train on a large unannotated CT image dataset and fine-tune on several CT classification tasks. Due to the challenge of obtaining sufficient annotated training data in medical imaging, it is of particular interest to evaluate how the self-supervised pre-training methods perform when fine-tuning on small datasets. By experimenting with gradually reducing the training dataset size for fine-tuning, we find that the reduction has different effects depending on the type of pre-training chosen. The SparK pre-training method is more robust to the training dataset size than the contrastive methods. Based on our results, we propose the SparK pre-training for medical imaging tasks with only small annotated datasets.
Nearest neighbor (NN) sampling provides more semantic variations than pre-defined transformations for self-supervised learning (SSL) based image recognition problems. However, its performance is restricted by the quality of the support set, which holds positive samples for the contrastive loss. In this work, we show that the quality of the support set plays a crucial role in any nearest neighbor based method for SSL. We then provide a refined baseline (pNNCLR) to the nearest neighbor based SSL approach (NNCLR). To this end, we introduce pseudo nearest neighbors (pNN) to control the quality of the support set, wherein, rather than sampling the nearest neighbors, we sample in the vicinity of hard nearest neighbors by varying the magnitude of the resultant vector and employing a stochastic sampling strategy to improve the performance. Additionally, to stabilize the effects of uncertainty in NN-based learning, we employ a smooth-weight-update approach for training the proposed network. Evaluation of the proposed method on multiple public image recognition and medical image recognition datasets shows that it performs up to 8 percent better than the baseline nearest neighbor method, and is comparable to other previously proposed SSL methods.
Event-based motion deblurring has shown promising results by exploiting low-latency events. However, current approaches are limited in their practical usage, as they assume the same spatial resolution of inputs and specific blurriness distributions. This work addresses these limitations and aims to generalize the performance of event-based deblurring in real-world scenarios. We propose a scale-aware network that allows flexible input spatial scales and enables learning from different temporal scales of motion blur. A two-stage self-supervised learning scheme is then developed to fit real-world data distribution. By utilizing the relativity of blurriness, our approach efficiently ensures the restored brightness and structure of latent images and further generalizes deblurring performance to handle varying spatial and temporal scales of motion blur in a self-distillation manner. Our method is extensively evaluated, demonstrating remarkable performance, and we also introduce a real-world dataset consisting of multi-scale blurry frames and events to facilitate research in event-based deblurring.
Classifying fine-grained lesions is challenging due to minor and subtle differences in medical images. This is because learning features of fine-grained lesions with highly minor differences is very difficult in training deep neural networks. Therefore, in this paper, we introduce Fine-Grained Self-Supervised Learning(FG-SSL) method for classifying subtle lesions in medical images. The proposed method progressively learns the model through hierarchical block such that the cross-correlation between the fine-grained Jigsaw puzzle and regularized original images is close to the identity matrix. We also apply hierarchical block for progressive fine-grained learning, which extracts different information in each step, to supervised learning for discovering subtle differences. Our method does not require an asymmetric model, nor does a negative sampling strategy, and is not sensitive to batch size. We evaluate the proposed fine-grained self-supervised learning method on comprehensive experiments using various medical image recognition datasets. In our experiments, the proposed method performs favorably compared to existing state-of-the-art approaches on the widely-used ISIC2018, APTOS2019, and ISIC2017 datasets.
Self-supervised learning (SSL) has gained significant interest in recent years as a solution to address the challenges posed by sparse and noisy data in recommender systems. Despite the growing number of SSL algorithms designed to provide state-of-the-art performance in various recommendation scenarios (e.g., graph collaborative filtering, sequential recommendation, social recommendation, KG-enhanced recommendation), there is still a lack of unified frameworks that integrate recommendation algorithms across different domains. Such a framework could serve as the cornerstone for self-supervised recommendation algorithms, unifying the validation of existing methods and driving the design of new ones. To address this gap, we introduce SSLRec, a novel benchmark platform that provides a standardized, flexible, and comprehensive framework for evaluating various SSL-enhanced recommenders. The SSLRec framework features a modular architecture that allows users to easily evaluate state-of-the-art models and a complete set of data augmentation and self-supervised toolkits to help create SSL recommendation models with specific needs. Furthermore, SSLRec simplifies the process of training and evaluating different recommendation models with consistent and fair settings. Our SSLRec platform covers a comprehensive set of state-of-the-art SSL-enhanced recommendation models across different scenarios, enabling researchers to evaluate these cutting-edge models and drive further innovation in the field. Our implemented SSLRec framework is available at the source code repository https://github.com/HKUDS/SSLRec.
Relevance modeling aims to locate desirable items for corresponding queries, which is crucial for search engines to ensure user experience. Although most conventional approaches address this problem by assessing the semantic similarity between the query and item, pure semantic matching is not everything. In reality, auxiliary query-item interactions extracted from user historical behavior data of the search log could provide hints to reveal users' search intents further. Drawing inspiration from this, we devise a novel Behavior Augmented Relevance Learning model for Alipay Search (BARL-ASe) that leverages neighbor queries of target item and neighbor items of target query to complement target query-item semantic matching. Specifically, our model builds multi-level co-attention for distilling coarse-grained and fine-grained semantic representations from both neighbor and target views. The model subsequently employs neighbor-target self-supervised learning to improve the accuracy and robustness of BARL-ASe by strengthening representation and logit learning. Furthermore, we discuss how to deal with the long-tail query-item matching of the mini apps search scenario of Alipay practically. Experiments on real-world industry data and online A/B testing demonstrate our proposal achieves promising performance with low latency.
Online groups have become increasingly prevalent, providing users with space to share experiences and explore interests. Therefore, user-centric group discovery task, i.e., recommending groups to users can help both users' online experiences and platforms' long-term developments. Existing recommender methods can not deal with this task as modeling user-group participation into a bipartite graph overlooks their item-side interests. Although there exist a few works attempting to address this task, they still fall short in fully preserving the social context and ensuring effective interest representation learning. In this paper, we focus on exploring the intents that motivate users to participate in groups, which can be categorized into different types, like the social-intent and the personal interest-intent. The former refers to users joining a group affected by their social links, while the latter relates to users joining groups with like-minded people for self-enjoyment. To comprehend different intents, we propose a novel model, DiRec, that first models each intent separately and then fuses them together for predictions. Specifically, for social-intent, we introduce the hypergraph structure to model the relationship between groups and members, leading to a richer understanding of the social context. As for interest-intent, we employ novel structural refinement on the interactive graph to uncover more intricate user behaviors and group interests, realizing better representation learning of interests. Furthermore, we also observe the intent overlapping in real-world scenarios and devise a novel self-supervised learning loss that encourages such alignment for final recommendations. Extensive experiments on three public datasets show the significant improvement of DiRec over the state-of-the-art methods.
Diffusion MRI is commonly performed using echo-planar imaging (EPI) due to its rapid acquisition time. However, the resolution of diffusion-weighted images is often limited by magnetic field inhomogeneity-related artifacts and blurring induced by T2- and T2*-relaxation effects. To address these limitations, multi-shot EPI (msEPI) combined with parallel imaging techniques is frequently employed. Nevertheless, reconstructing msEPI can be challenging due to phase variation between multiple shots. In this study, we introduce a novel msEPI reconstruction approach called zero-MIRID (zero-shot self-supervised learning of Multi-shot Image Reconstruction for Improved Diffusion MRI). This method jointly reconstructs msEPI data by incorporating deep learning-based image regularization techniques. The network incorporates CNN denoisers in both k- and image-spaces, while leveraging virtual coils to enhance image reconstruction conditioning. By employing a self-supervised learning technique and dividing sampled data into three groups, the proposed approach achieves superior results compared to the state-of-the-art parallel imaging method, as demonstrated in an in-vivo experiment.
Invariance against rotations of 3D objects is an important property in analyzing 3D point set data. Conventional 3D point set DNNs having rotation invariance typically obtain accurate 3D shape features via supervised learning by using labeled 3D point sets as training samples. However, due to the rapid increase in 3D point set data and the high cost of labeling, a framework to learn rotation-invariant 3D shape features from numerous unlabeled 3D point sets is required. This paper proposes a novel self-supervised learning framework for acquiring accurate and rotation-invariant 3D point set features at object-level. Our proposed lightweight DNN architecture decomposes an input 3D point set into multiple global-scale regions, called tokens, that preserve the spatial layout of partial shapes composing the 3D object. We employ a self-attention mechanism to refine the tokens and aggregate them into an expressive rotation-invariant feature per 3D point set. Our DNN is effectively trained by using pseudo-labels generated by a self-distillation framework. To facilitate the learning of accurate features, we propose to combine multi-crop and cut-mix data augmentation techniques to diversify 3D point sets for training. Through a comprehensive evaluation, we empirically demonstrate that, (1) existing rotation-invariant DNN architectures designed for supervised learning do not necessarily learn accurate 3D shape features under a self-supervised learning scenario, and (2) our proposed algorithm learns rotation-invariant 3D point set features that are more accurate than those learned by existing algorithms. Code is available at https://github.com/takahikof/RIPT_SDMM
Noisy labels hurt deep learning-based supervised image classification performance as the models may overfit the noise and learn corrupted feature extractors. For natural image classification training with noisy labeled data, model initialization with contrastive self-supervised pretrained weights has shown to reduce feature corruption and improve classification performance. However, no works have explored: i) how other self-supervised approaches, such as pretext task-based pretraining, impact the learning with noisy label, and ii) any self-supervised pretraining methods alone for medical images in noisy label settings. Medical images often feature smaller datasets and subtle inter class variations, requiring human expertise to ensure correct classification. Thus, it is not clear if the methods improving learning with noisy labels in natural image datasets such as CIFAR would also help with medical images. In this work, we explore contrastive and pretext task-based self-supervised pretraining to initialize the weights of a deep learning classification model for two medical datasets with self-induced noisy labels -- NCT-CRC-HE-100K tissue histological images and COVID-QU-Ex chest X-ray images. Our results show that models initialized with pretrained weights obtained from self-supervised learning can effectively learn better features and improve robustness against noisy labels.