Dense Semantic Contrast for Self-Supervised Visual Representation Learning

Self-supervised learning (SSL) is a challenging task in sleep stage classification (SSC) that is capable of mining valuable representations from unlabeled data. However, traditional SSL methods typically focus on single-view learning and do not fully exploit the interactions among information across multiple views. In this study, we focused on a multi-domain view of the same EEG signal and developed a self-supervised multi-view representation learning framework via time series and time-frequency contrasting (MV-TTFC). In the MV-TTFC framework, we built-in a cross-domain view contrastive learning prediction task to establish connections between the temporal view and time-frequency (TF) view, thereby enhancing the information exchange between multiple views. In addition, to improve the quality of the TF view inputs, we introduced an enhanced multisynchrosqueezing transform, which can create high energy concentration TF image views to compensate for the inaccurate representations in traditional TF processing techniques. Finally, integrating temporal, TF, and fusion space contrastive learning effectively captured the latent features in EEG signals. We evaluated MV-TTFC based on two real-world SSC datasets (SleepEDF-78 and SHHS) and compared it with baseline methods in downstream tasks. Our method exhibited state-of-the-art performance, achieving accuracies of 78.64% and 81.45% with SleepEDF-78 and SHHS, respectively, and macro F1-scores of 70.39% with SleepEDF-78 and 70.47% with SHHS.

Prompting techniques unlock valuable insights from pre-trained models, leading to a remarkable performance in Natural Language Processing (NLP). Inspired by NLP advancements, various graph prompt methods have emerged to transfer knowledge from pre-trained graph models to diverse downstream graph tasks. However, designing appropriate prompts for different downstream tasks remains a significant challenge due to substantial differences between pre-training graph tasks and downstream graph tasks. Moreover, existing graph prompt methods prevent transferability across different domains as they independently pre-train a model for each dataset. To address these challenges, we propose Graph Pretraining and Task-Specific Prompting (GTSP). To the best of our knowledge, GTSP is the first unified pre-training framework to handle cross-domain graph data and multiple types of graph tasks simultaneously. The key idea of our framework is to utilize similar label features between pre-training tasks and downstream tasks to seamlessly integrate the downstream task prompt into the pretrained model. We then propose a task-specific prompting architecture and design a label similarity matrix as prompt input. With GTSP, the pre-trained model can adapt to various types of downstream graph tasks, effectively avoiding the challenges of forcibly altering graph task forms. Experimental analysis conducted on benchmark datasets demonstrates the superiority of our method.

Traditional approaches for learning representations from somatic mutation data often rely on the distribution of motif counts, which have limitations, including the use of small sequence contexts and the exclusion of critical information, such as mutation location and interactions among mutations. Recent efforts have leveraged advances in natural language processing (NLP) to improve representation learning from somatic mutations by incorporating larger sequence contexts into shallow networks. More recently, studies have demonstrated the feasibility of leveraging attention mechanisms to learn meaningful representations. However, these methods primarily use supervised or weakly supervised learning paradigms, which are inherently task-specific. The development of general-purpose representations, by harnessing the vast amount of somatic mutation data through self-supervised techniques, remains underexplored. Such representations could drive progress in applications including tumor subtype classification, biomarker discovery, patient stratification, drug response prediction, and pathway enrichment analysis. This study aims to develop a deep, unsupervised learning framework that leverages abundant somatic mutation data to generate robust, general-purpose representations. By adapting recent advances in self-supervised learning, the proposed approach seeks to establish a foundation model capable of capturing meaningful patterns in mutation data for diverse downstream tasks. The framework employs novel tokenization and embedding strategies to encode mutation sequences and utilizes transformer-based architectures to model contextual dependencies. The process of inputting mutation data into a transformer architecture, particularly the tokenization step, is critical for this problem, and we explore various tokenization strategies. Benchmarking these representations will include evaluating their utility across tumor subtype classification, biomarker discovery, patient stratification, drug response prediction, identification of mutational signatures. Although the model is under development, experiments from prior work—limited to mutation identifiers without leveraging sequence motifs, mutation location, or other contextual information—have demonstrated the feasibility of encoding mutational data into meaningful representations using transformer-based architectures and self-supervised learning. Building on these findings, this framework will benchmark its performance on multiple downstream tasks, including tumor subtype classification and patient stratification. Citation Format: Marzieh Haghighi, Gustavo Alonso Arango Argoty, Etai Jacob. Self-supervised representation learning of somatic mutational data [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 5059.

MotivationProtein representation learning methods have shown great potential to many downstream tasks in biological applications. A few recent studies have demonstrated that the self-supervised learning is a promising solution to addressing insufficient labels of proteins, which is a major obstacle to effective protein representation learning. However, existing protein representation learning is usually pretrained on protein sequences without considering the important protein structural information.ResultsIn this work, we propose a novel structure-aware protein self-supervised learning method to effectively capture structural information of proteins. In particular, a graph neural network model is pretrained to preserve the protein structural information with self-supervised tasks from a pairwise residue distance perspective and a dihedral angle perspective, respectively. Furthermore, we propose to leverage the available protein language model pretrained on protein sequences to enhance the self-supervised learning. Specifically, we identify the relation between the sequential information in the protein language model and the structural information in the specially designed graph neural network model via a novel pseudo bi-level optimization scheme. We conduct experiments on three downstream tasks: the binary classification into membrane/non-membrane proteins, the location classification into 10 cellular compartments, and the enzyme-catalyzed reaction classification into 384 EC numbers, and these experiments verify the effectiveness of our proposed method.Availability and implementationThe Alphafold2 database is available in https://alphafold.ebi.ac.uk/. The PDB files are available in https://www.rcsb.org/. The downstream tasks are available in https://github.com/phermosilla/IEConv\\_proteins/tree/master/Datasets. The code of the proposed method is available in https://github.com/GGchen1997/STEPS_Bioinformatics.

To date, various 3D scene understanding tasks still lack practical and generalizable pre-trained models, primarily due to the intricate nature of 3D scene understanding tasks and their immense variations introduced by camera views, lighting, occlusions, etc. In this paper, we tackle this challenge by introducing a spatio-temporal representation learning (STRL) framework, capable of learning from unlabeled 3D point clouds in a self-supervised fashion. Inspired by how infants learn from visual data in the wild, we explore the rich spatio-temporal cues derived from the 3D data. Specifically, STRL takes two temporally-correlated frames from a 3D point cloud sequence as the input, transforms it with the spatial data augmentation, and learns the invariant representation self-supervisedly. To corroborate the efficacy of STRL, we conduct extensive experiments on three types (synthetic, indoor, and outdoor) of datasets. Experimental results demonstrate that, compared with supervised learning methods, the learned self-supervised representation facilitates various models to attain comparable or even better performances while capable of generalizing pre-trained models to downstream tasks, including 3D shape classification, 3D object detection, and 3D semantic segmentation. Moreover, the spatio-temporal contextual cues embedded in 3D point clouds significantly improve the learned representations.

Panoptic Segmentation aims to provide an understanding of background (stuff) and instances of objects (things) at a pixel level. It combines the separate tasks of semantic segmentation (pixel level classification) and instance segmentation to build a single unified scene understanding task. Typically, panoptic segmentation is derived by combining semantic and instance segmentation tasks that are learned separately or jointly (multi-task networks). In general, instance segmentation networks are built by adding a foreground mask estimation layer on top of object detectors or using instance clustering methods that assign a pixel to an instance center. In this work, we present a fully convolution neural network that learns instance segmentation from semantic segmentation and instance contours (boundaries of things). Instance contours along with semantic segmentation yield a boundary aware semantic segmentation of things. Connected component labeling on these results produces instance segmentation. We merge semantic and instance segmentation results to output panoptic segmentation. We evaluate our proposed method on the CityScapes dataset to demonstrate qualitative and quantitative performances along with several ablation studies. Our overview video can be accessed from https://youtu.be/wBtcxRhG3e0.

DrasCLR: A self-supervised framework of learning disease-related and anatomy-specific representation for 3D lung CT images

Attribute graphs are a crucial data structure for graph communities. However, the presence of redundancy and noise in the attribute graph can impair the aggregation effect of integrating two different heterogeneous distributions of attribute and structural features, resulting in inconsistent and distorted data that ultimately compromises the accuracy and reliability of attribute graph learning. For instance, redundant or irrelevant attributes can result in overfitting, while noisy attributes can lead to underfitting. Similarly, redundant or noisy structural features can affect the accuracy of graph representations, making it challenging to distinguish between different nodes or communities. To address these issues, we propose the embedded fusion graph auto-encoder framework for self-supervised learning (SSL), which leverages multitask learning to fuse node features across different tasks to reduce redundancy. The embedding fusion graph auto-encoder (EFGAE) framework comprises two phases: pretraining (PT) and downstream task learning (DTL). During the PT phase, EFGAE uses a graph auto-encoder (GAE) based on adversarial contrastive learning to learn structural and attribute embeddings separately and then fuses these embeddings to obtain a representation of the entire graph. During the DTL phase, we introduce an adaptive graph convolutional network (AGCN), which is applied to graph neural network (GNN) classifiers to enhance recognition for downstream tasks. The experimental results demonstrate that our approach outperforms state-of-the-art (SOTA) techniques in terms of accuracy, generalization ability, and robustness.

Unsupervised/self-supervised pre-training methods for graph representation learning have recently attracted increasing research interests, and they are shown to be able to generalize to various downstream applications. Yet, the adversarial robustness of such pre-trained graph learning models remains largely unexplored. More importantly, most existing defense techniques designed for end-to-end graph representation learning methods require pre-specified label definitions, and thus cannot be directly applied to the pre-training methods. In this paper, we propose an unsupervised defense technique to robustify pre-trained deep graph models, so that the perturbations on the input graph can be successfully identified and blocked before the model is applied to different downstream tasks. Specifically, we introduce a mutual information-based measure, graph representation vulnerability (GRV), to quantify the robustness of graph encoders on the representation space. We then formulate an optimization problem to learn the graph representation by carefully balancing the trade-off between the expressive power and the robustness (i.e., GRV) of the graph encoder. The discrete nature of graph topology and the joint space of graph data make the optimization problem intractable to solve. To handle the above difficulty and to reduce computational expense, we further relax the problem and thus provide an approximate solution. Additionally, we explore a provable connection between the robustness of the unsupervised graph encoder and that of models on downstream tasks. Extensive experiments demonstrate that even without access to labels and tasks, our model is still able to enhance robustness against adversarial attacks on three downstream tasks (node classification, link prediction, and community detection) by an average of +16.5% compared with existing methods.

Self-supervised learning has become the cornerstone of building generalizable and transferable artificial intelligence systems in medical imaging. In particular, contrastive representation learning techniques trained on large multi-modal datasets have demonstrated impressive capabilities of producing highly transferable representations for different downstream tasks. In ophthalmology, large multi-modal datasets are abundantly available and conveniently accessible as modern retinal imaging scanners acquire both 2D fundus images and 3D optical coherence tomography (OCT) scans to assess the eye. In this context, we introduce a novel multi-modal contrastive learning-based pipeline to facilitate learning joint representations for the two retinal imaging modalities. After self-supervised pre-training on 153,306 scan pairs, we show that such a pre-training framework can provide both a retrieval system and encoders that produce comprehensive OCT and fundus image representations that generalize well for various downstream tasks on three independent external datasets, explicitly focusing on clinically pertinent prediction tasks. In addition, we show that interchanging OCT with lower-cost fundus imaging can preserve the predictive power of the trained models.

Unsupervised and self-supervised representation learning has become popular in recent years for learning useful features from unlabelled data. Representation learning has been mostly developed in the neural network literature, and other models for representation learning are surprisingly unexplored. In this work, we introduce and analyze several kernel-based representation learning approaches: Firstly, we define two kernel Self-Supervised Learning (SSL) models using contrastive loss functions and secondly, a Kernel Autoencoder (AE) model based on the idea of embedding and reconstructing data. We argue that the classical representer theorems for supervised kernel machines are not always applicable for (self-supervised) representation learning, and present new representer theorems, which show that the representations learned by our kernel models can be expressed in terms of kernel matrices. We further derive generalisation error bounds for representation learning with kernel SSL and AE, and empirically evaluate the performance of these methods in both small data regimes as well as in comparison with neural network based models.

Self-supervised learning (SSL) has achieved great success in speech-related tasks. While Transformer and Conformer architectures have dominated SSL backbones, encoders like Zipformer, which excel in automatic speech recognition (ASR), remain unexplored in SSL. Concurrently, inefficiencies in data processing within existing SSL training frameworks, such as fairseq, pose challenges in managing the growing volumes of training data. To address these issues, we propose k2SSL, an open-source framework that offers faster, more memory-efficient, and better-performing self-supervised speech representation learning, focusing on downstream ASR tasks. The optimized HuBERT and proposed Zipformer-based SSL systems exhibit substantial reductions in both training time and memory usage during SSL training. Experiments on LibriSpeech demonstrate that Zipformer Base significantly outperforms HuBERT and WavLM, achieving up to a 34.8% relative WER reduction compared to HuBERT Base after fine-tuning, along with a 3.5x pre-training speedup in GPU hours. When scaled to 60k hours of LibriLight data, Zipformer Large exhibits remarkable efficiency, matching HuBERT Large’s performance while requiring only 5/8 pre-training steps.

Depression detection research has increased over the last few decades, one major bottleneck of which is the limited data availability and representation learning. Recently, self-supervised learning has seen success in pretraining text embeddings and has been applied broadly on related tasks with sparse data, while pretrained audio embeddings based on self-supervised learning are rarely investigated. This paper proposes DEPA, a self-supervised, pretrained dep ression a udio embedding method for depression detection. An encoder-decoder network is used to extract DEPA on in-domain depressed datasets (DAIC and MDD) and out-domain (Switchboard, Alzheimer's) datasets. With DEPA as the audio embedding extracted at response-level, a significant performance gain is achieved on downstream tasks, evaluated on both sparse datasets like DAIC and large major depression disorder dataset (MDD). This paper not only exhibits itself as a novel embedding extracting method capturing response-level representation for depression detection but more significantly, is an exploration of self-supervised learning in a specific task within audio processing.

Building behavior profiles of Android applications (apps) with holistic, rich and multi-view information (e.g., incorporating several semantic views of an app such as API sequences, system calls, etc.) would help catering downstream analytics tasks such as app categorization, recommendation and malware analysis significantly better. Towards this goal, we design a semi-supervised Representation Learning (RL) framework named apk2vec to automatically generate a compact representation (aka profile/embedding) for a given app. More specifically, apk2vec has the three following unique characteristics which make it an excellent choice for largescale app profiling: (1) it encompasses information from multiple semantic views such as API sequences, permissions, etc., (2) being a semi-supervised embedding technique, it can make use of labels associated with apps (e.g., malware family or app category labels) to build high quality app profiles, and (3) it combines RL and feature hashing which allows it to efficiently build profiles of apps that stream over time (i.e., online learning). The resulting semi-supervised multi-view hash embeddings of apps could then be used for a wide variety of downstream tasks such as the ones mentioned above. Our extensive evaluations with more than 42,000 apps demonstrate that apk2vec's app profiles could significantly outperform state-of-the-art techniques in four app analytics tasks namely, malware detection, familial clustering, app clone detection and app recommendation.

Learning meaningful representations for time series is challenging when labeled data are scarce or entirely unavailable. Self-supervised contrastive learning has recently emerged as an effective approach to extract informative features from unlabeled data by contrasting multiple augmented views. In this study, we introduce TS-TCC, a novel framework for time-series representation learning that leverages contrastive learning on unlabeled sequences. Our approach designs time-series-specific weak and strong augmentations and employs them in a temporal contrasting module to capture robust temporal dependencies, while a contextual contrasting module enhances the discriminative quality of the learned features. We also conduct a comprehensive analysis of augmentation strategies, which are crucial for enhancing the effectiveness of contrastive learning. Furthermore, we extend TS-TCC to semi-supervised settings with Class-Aware TS-TCC (CA-TCC), which utilizes limited labeled data to refine representations by incorporating a class-aware contrastive loss based on high-quality pseudo labels generated by TS-TCC. Extensive experiments demonstrate that the learned features achieve linear evaluation performance comparable to fully supervised training and exhibit strong effectiveness in scenarios with few labels and in transfer learning tasks.