Contrastive self-supervised learning (CSL) with a prototypical regularization has been introduced in learning meaningful representations for downstream tasks that require strong semantic information. However, to optimize CSL with a loss that performs the prototypical regularization aggressively, e.g., the ProtoNCE loss, might cause the "coagulation" of examples in the embedding space. That is, the intra-prototype diversity of samples collapses to trivial solutions for their prototype being well-separated from others. Motivated by previous works, we propose to mitigate this phenomenon by learning Prototypical representation through Alignment, Uniformity and Correlation (PAUC). Specifically, the ordinary ProtoNCE loss is revised with: (1) an alignment loss that pulls embeddings from positive prototypes together; (2) a uniformity loss that distributes the prototypical level features uniformly; (3) a correlation loss that increases the diversity and discriminability between prototypical level features. We conduct extensive experiments on various benchmarks where the results demonstrate the effectiveness of our method in improving the quality of prototypical contrastive representations. Particularly, in the classification down-stream tasks with linear probes, our proposed method outperforms the state-of-the-art instance-wise and prototypical contrastive learning methods on the ImageNet-100 dataset by 2.96% and the ImageNet-1K dataset by 2.46% under the same settings of batch size and epochs.
Existing vision-text contrastive learning like CLIP aims to match the paired image and caption embeddings while pushing others apart, which improves representation transferability and supports zero-shot prediction. However, medical image-text datasets are orders of magnitude below the general images and captions from the internet. Moreover, previous methods encounter many false negatives, i.e., images and reports from separate patients probably carry the same semantics but are wrongly treated as negatives. In this paper, we decouple images and texts for multimodal contrastive learning thus scaling the usable training data in a combinatorial magnitude with low cost. We also propose to replace the InfoNCE loss with semantic matching loss based on medical knowledge to eliminate false negatives in contrastive learning. We prove that MedCLIP is a simple yet effective framework: it outperforms state-of-the-art methods on zero-shot prediction, supervised classification, and image-text retrieval. Surprisingly, we observe that with only 20K pre-training data, MedCLIP wins over the state-of-the-art method (using around 200K data). Our code is available at https://github.com/RyanWangZf/MedCLIP.
Unsupervised representation learning for dynamic graphs has attracted a lot of research attention in recent years. Compared with static graph, the dynamic graph is a comprehensive embodiment of both the intrinsic stable characteristics of nodes and the time-related dynamic preference. However, existing methods generally mix these two types of information into a single representation space, which may lead to poor explanation, less robustness, and a limited ability when applied to different downstream tasks. To solve the above problems, in this paper, we propose a novel disenTangled representation learning framework for discrete-time Dynamic graphs, namely DyTed. We specially design a temporal-clips contrastive learning task together with a structure contrastive learning to effectively identify the time-invariant and time-varying representations respectively. To further enhance the disentanglement of these two types of representation, we propose a disentanglement-aware discriminator under an adversarial learning framework from the perspective of information theory. Extensive experiments on Tencent and five commonly used public datasets demonstrate that DyTed, as a general framework that can be applied to existing methods, achieves state-of-the-art performance on various downstream tasks, as well as be more robust against noise.
Visualization methods based on the nearest neighbor graph, such as t-SNE or UMAP, are widely used for visualizing high-dimensional data. Yet, these approaches only produce meaningful results if the nearest neighbors themselves are meaningful. For images represented in pixel space this is not the case, as distances in pixel space are often not capturing our sense of similarity and therefore neighbors are not semantically close. This problem can be circumvented by self-supervised approaches based on contrastive learning, such as SimCLR, relying on data augmentation to generate implicit neighbors, but these methods do not produce two-dimensional embeddings suitable for visualization. Here, we present a new method, called t-SimCNE, for unsupervised visualization of image data. T-SimCNE combines ideas from contrastive learning and neighbor embeddings, and trains a parametric mapping from the high-dimensional pixel space into two dimensions. We show that the resulting 2D embeddings achieve classification accuracy comparable to the state-of-the-art high-dimensional SimCLR representations, thus faithfully capturing semantic relationships. Using t-SimCNE, we obtain informative visualizations of the CIFAR-10 and CIFAR-100 datasets, showing rich cluster structure and highlighting artifacts and outliers.
In this paper, we describe a universal method for extracting the underlying monotonic trend factor from time series data. We propose an approach related to the Mann-Kendall test, a standard monotonic trend detection method and call it contrastive trend estimation (CTE). We show that the CTE method identifies any hidden trend underlying temporal data while avoiding the standard assumptions used for monotonic trend identification. In particular, CTE can take any type of temporal data (vector, images, graphs, time series, etc.) as input. We finally illustrate the interest of our CTE method through several experiments on different types of data and problems.
Most existing pre-trained language representation models (PLMs) are sub-optimal in sentiment analysis tasks, as they capture the sentiment information from word-level while under-considering sentence-level information. In this paper, we propose SentiWSP, a novel Sentiment-aware pre-trained language model with combined Word-level and Sentence-level Pre-training tasks. The word level pre-training task detects replaced sentiment words, via a generator-discriminator framework, to enhance the PLM's knowledge about sentiment words. The sentence level pre-training task further strengthens the discriminator via a contrastive learning framework, with similar sentences as negative samples, to encode sentiments in a sentence. Extensive experimental results show that SentiWSP achieves new state-of-the-art performance on various sentence-level and aspect-level sentiment classification benchmarks. We have made our code and model publicly available at https://github.com/XMUDM/SentiWSP.
Learning representations for individual instances when only bag-level labels are available is a fundamental challenge in multiple instance learning (MIL). Recent works have shown promising results using contrastive self-supervised learning (CSSL), which learns to push apart representations corresponding to two different randomly-selected instances. Unfortunately, in real-world applications such as medical image classification, there is often class imbalance, so randomly-selected instances mostly belong to the same majority class, which precludes CSSL from learning inter-class differences. To address this issue, we propose a novel framework, Iterative Self-paced Supervised Contrastive Learning for MIL Representations (ItS2CLR), which improves the learned representation by exploiting instance-level pseudo labels derived from the bag-level labels. The framework employs a novel self-paced sampling strategy to ensure the accuracy of pseudo labels. We evaluate ItS2CLR on three medical datasets, showing that it improves the quality of instance-level pseudo labels and representations, and outperforms existing MIL methods in terms of both bag and instance level accuracy. Code is available at https://github.com/Kangningthu/ItS2CLR
Knowledge tracing and dropout prediction are crucial for online education to estimate students' knowledge states or to prevent dropout rates. While traditional systems interacting with students suffered from data sparsity and overfitting, recent sample-level contrastive learning helps to alleviate this issue. One major limitation of sample-level approaches is that they regard students' behavior interaction sequences as a bundle, so they often fail to encode temporal contexts and track their dynamic changes, making it hard to find optimal representations for knowledge tracing and dropout prediction. To apply temporal context within the sequence, this study introduces a novel student modeling framework, SAICL: \textbf{s}tudent modeling with \textbf{a}uxiliary \textbf{i}nteraction-level \textbf{c}ontrastive \textbf{l}earning. In detail, SAICL can utilize both proposed self-supervised/supervised interaction-level contrastive objectives: MilCPC (\textbf{M}ulti-\textbf{I}nteraction-\textbf{L}evel \textbf{C}ontrastive \textbf{P}redictive \textbf{C}oding) and SupCPC (\textbf{Sup}ervised \textbf{C}ontrastive \textbf{P}redictive \textbf{C}oding). While previous sample-level contrastive methods for student modeling are highly dependent on data augmentation methods, the SAICL is free of data augmentation while showing better performance in both self-supervised and supervised settings. By combining cross-entropy with contrastive objectives, the proposed SAICL achieved comparable knowledge tracing and dropout prediction performance with other state-of-art models without compromising inference costs.
An ideal audio retrieval system efficiently and robustly recognizes a short query snippet from an extensive database. However, the performance of well-known audio fingerprinting systems falls short at high signal distortion levels. This paper presents an audio retrieval system that generates noise and reverberation robust audio fingerprints using the contrastive learning framework. Using these fingerprints, the method performs a comprehensive search to identify the query audio and precisely estimate its timestamp in the reference audio. Our framework involves training a CNN to maximize the similarity between pairs of embeddings extracted from clean audio and its corresponding distorted and time-shifted version. We employ a channel-wise spectral-temporal attention mechanism to better discriminate the audio by giving more weight to the salient spectral-temporal patches in the signal. Experimental results indicate that our system is efficient in computation and memory usage while being more accurate, particularly at higher distortion levels, than competing state-of-the-art systems and scalable to a larger database.
Traditional end-to-end task-oriented dialog systems first convert dialog context into belief state and action state before generating the system response. The system response performance is significantly affected by the quality of the belief state and action state. We first explore what dialog context representation is beneficial to improving the quality of the belief state and action state, which further enhances the generated response quality. To tackle our exploration, we propose Mars, an end-to-end task-oriented dialog system with two contrastive learning strategies to model the relationship between dialog context and belief/action state representations. Empirical results show dialog context representations, which are more different from semantic state representations, are more conducive to multi-turn task-oriented dialog. Moreover, our proposed Mars achieves state-of-the-art performance on the MultiWOZ 2.0, CamRest676, and CrossWOZ.
Discovering out-of-domain (OOD) intent is important for developing new skills in task-oriented dialogue systems. The key challenges lie in how to transfer prior in-domain (IND) knowledge to OOD clustering, as well as jointly learn OOD representations and cluster assignments. Previous methods suffer from in-domain overfitting problem, and there is a natural gap between representation learning and clustering objectives. In this paper, we propose a unified K-nearest neighbor contrastive learning framework to discover OOD intents. Specifically, for IND pre-training stage, we propose a KCL objective to learn inter-class discriminative features, while maintaining intra-class diversity, which alleviates the in-domain overfitting problem. For OOD clustering stage, we propose a KCC method to form compact clusters by mining true hard negative samples, which bridges the gap between clustering and representation learning. Extensive experiments on three benchmark datasets show that our method achieves substantial improvements over the state-of-the-art methods.
Conventional event detection models under supervised learning settings suffer from the inability of transfer to newly-emerged event types owing to lack of sufficient annotations. A commonly-adapted solution is to follow a identify-then-classify manner, which first identifies the triggers and then converts the classification task via a few-shot learning paradigm. However, these methods still fall far short of expectations due to: (i) insufficient learning of discriminative representations in low-resource scenarios, and (ii) trigger misidentification caused by the overlap of the learned representations of triggers and non-triggers. To address the problems, in this paper, we propose a novel Hybrid Contrastive Learning method with a Task-Adaptive Threshold (abbreviated as HCLTAT), which enables discriminative representation learning with a two-view contrastive loss (support-support and prototype-query), and devises a easily-adapted threshold to alleviate misidentification of triggers. Extensive experiments on the benchmark dataset FewEvent demonstrate the superiority of our method to achieve better results compared to the state-of-the-arts. All the code and data of this paper will be available for online public access.
Graph contrastive learning (GCL) has recently emerged as an effective learning paradigm to alleviate the reliance on labelling information for graph representation learning. The core of GCL is to maximise the mutual information between the representation of a node and its contextual representation (i.e., the corresponding instance with similar semantic information) summarised from the contextual scope (e.g., the whole graph or 1-hop neighbourhood). This scheme distils valuable self-supervision signals for GCL training. However, existing GCL methods still suffer from limitations, such as the incapacity or inconvenience in choosing a suitable contextual scope for different datasets and building biased contrastiveness. To address aforementioned problems, we present a simple self-supervised learning method termed Unifying Graph Contrastive Learning with Flexible Contextual Scopes (UGCL for short). Our algorithm builds flexible contextual representations with tunable contextual scopes by controlling the power of an adjacency matrix. Additionally, our method ensures contrastiveness is built within connected components to reduce the bias of contextual representations. Based on representations from both local and contextual scopes, UGCL optimises a very simple contrastive loss function for graph representation learning. Essentially, the architecture of UGCL can be considered as a general framework to unify existing GCL methods. We have conducted intensive experiments and achieved new state-of-the-art performance in six out of eight benchmark datasets compared with self-supervised graph representation learning baselines. Our code has been open-sourced.
Indoor self-localization is a highly demanded system function for smartphones. The current solutions based on inertial, radio frequency, and geomagnetic sensing may have degraded performance when their limiting factors take effect. In this paper, we present a new indoor simultaneous localization and mapping (SLAM) system that utilizes the smartphone's built-in audio hardware and inertial measurement unit (IMU). Our system uses a smartphone's loudspeaker to emit near-inaudible chirps and then the microphone to record the acoustic echoes from the indoor environment. Our profiling measurements show that the echoes carry location information with sub-meter granularity. To enable SLAM, we apply contrastive learning to construct an echoic location feature (ELF) extractor, such that the loop closures on the smartphone's trajectory can be accurately detected from the associated ELF trace. The detection results effectively regulate the IMU-based trajectory reconstruction. Extensive experiments show that our ELF-based SLAM achieves median localization errors of $0.1\,\text{m}$, $0.53\,\text{m}$, and $0.4\,\text{m}$ on the reconstructed trajectories in a living room, an office, and a shopping mall, and outperforms the Wi-Fi and geomagnetic SLAM systems.
Recently, dense contrastive learning has shown superior performance on dense prediction tasks compared to instance-level contrastive learning. Despite its supremacy, the properties of dense contrastive representations have not yet been carefully studied. Therefore, we analyze the theoretical ideas of dense contrastive learning using a standard CNN and straightforward feature matching scheme rather than propose a new complex method. Inspired by the analysis of the properties of instance-level contrastive representations through the lens of alignment and uniformity on the hypersphere, we employ and extend the same lens for the dense contrastive representations to analyze their underexplored properties. We discover the core principle in constructing a positive pair of dense features and empirically proved its validity. Also, we introduces a new scalar metric that summarizes the correlation between alignment-and-uniformity and downstream performance. Using this metric, we study various facets of densely learned contrastive representations such as how the correlation changes over single- and multi-object datasets or linear evaluation and dense prediction tasks. The source code is publicly available at: https://github.com/SuperSupermoon/DenseCL-analysis
Recent studies have demonstrated the superior performance of introducing ``scan-wise" contrast labels into contrastive learning for multi-organ segmentation on multi-phase computed tomography (CT). However, such scan-wise labels are limited: (1) a coarse classification, which could not capture the fine-grained ``organ-wise" contrast variations across all organs; (2) the label (i.e., contrast phase) is typically manually provided, which is error-prone and may introduce manual biases of defining phases. In this paper, we propose a novel data-driven contrastive loss function that adapts the similar/dissimilar contrast relationship between samples in each minibatch at organ-level. Specifically, as variable levels of contrast exist between organs, we hypothesis that the contrast differences in the organ-level can bring additional context for defining representations in the latent space. An organ-wise contrast correlation matrix is computed with mean organ intensities under one-hot attention maps. The goal of adapting the organ-driven correlation matrix is to model variable levels of feature separability at different phases. We evaluate our proposed approach on multi-organ segmentation with both non-contrast CT (NCCT) datasets and the MICCAI 2015 BTCV Challenge contrast-enhance CT (CECT) datasets. Compared to the state-of-the-art approaches, our proposed contrastive loss yields a substantial and significant improvement of 1.41% (from 0.923 to 0.936, p-value$<$0.01) and 2.02% (from 0.891 to 0.910, p-value$<$0.01) on mean Dice scores across all organs with respect to NCCT and CECT cohorts. We further assess the trained model performance with the MICCAI 2021 FLARE Challenge CECT datasets and achieve a substantial improvement of mean Dice score from 0.927 to 0.934 (p-value$<$0.01). The code is available at: https://github.com/MASILab/DCC_CL
Deep models have been widely and successfully used in image manipulation detection, which aims to classify tampered images and localize tampered regions. Most existing methods mainly focus on extracting global features from tampered images, while neglecting the relationships of local features between tampered and authentic regions within a single tampered image. To exploit such spatial relationships, we propose Proposal Contrastive Learning (PCL) for effective image manipulation detection. Our PCL consists of a two-stream architecture by extracting two types of global features from RGB and noise views respectively. To further improve the discriminative power, we exploit the relationships of local features through a proxy proposal contrastive learning task by attracting/repelling proposal-based positive/negative sample pairs. Moreover, we show that our PCL can be easily adapted to unlabeled data in practice, which can reduce manual labeling costs and promote more generalizable features. Extensive experiments among several standard datasets demonstrate that our PCL can be a general module to obtain consistent improvement. The code is available at https://github.com/Sandy-Zeng/PCL.
Masked Autoencoders (MAE) based on a reconstruction task have risen to be a promising paradigm for self-supervised learning (SSL) and achieve state-of-the-art performance across different benchmark datasets. However, despite its impressive empirical success, there is still limited theoretical understanding of it. In this paper, we propose a theoretical understanding of how masking matters for MAE to learn meaningful features. We establish a close connection between MAE and contrastive learning, which shows that MAE implicit aligns the mask-induced positive pairs. Built upon this connection, we develop the first downstream guarantees for MAE methods, and analyze the effect of mask ratio. Besides, as a result of the implicit alignment, we also point out the dimensional collapse issue of MAE, and propose a Uniformity-enhanced MAE (U-MAE) loss that can effectively address this issue and bring significant improvements on real-world datasets, including CIFAR-10, ImageNet-100, and ImageNet-1K. Code is available at (https://github.com/zhangq327/U-MAE).
Contrastive language-image pre-training (CLIP) serves as a de-facto standard to align images and texts. Nonetheless, the loose correlation between images and texts of web-crawled data renders the contrastive objective data inefficient and craving for a large training batch size. In this work, we explore the validity of non-contrastive language-image pre-training (nCLIP), and study whether nice properties exhibited in visual self-supervised models can emerge. We empirically observe that the non-contrastive objective nourishes representation learning while sufficiently underperforming under zero-shot recognition. Based on the above study, we further introduce xCLIP, a multi-tasking framework combining CLIP and nCLIP, and show that nCLIP aids CLIP in enhancing feature semantics. The synergy between two objectives lets xCLIP enjoy the best of both worlds: superior performance in both zero-shot transfer and representation learning. Systematic evaluation is conducted spanning a wide variety of downstream tasks including zero-shot classification, out-of-domain classification, retrieval, visual representation learning, and textual representation learning, showcasing a consistent performance gain and validating the effectiveness of xCLIP.
We revisit the weakly supervised cross-modal face-name alignment task; that is, given an image and a caption, we label the faces in the image with the names occurring in the caption. Whereas past approaches have learned the latent alignment between names and faces by uncertainty reasoning over a set of images and their respective captions, in this paper, we rely on appropriate loss functions to learn the alignments in a neural network setting and propose SECLA and SECLA-B. SECLA is a Symmetry-Enhanced Contrastive Learning-based Alignment model that can effectively maximize the similarity scores between corresponding faces and names in a weakly supervised fashion. A variation of the model, SECLA-B, learns to align names and faces as humans do, that is, learning from easy to hard cases to further increase the performance of SECLA. More specifically, SECLA-B applies a two-stage learning framework: (1) Training the model on an easy subset with a few names and faces in each image-caption pair. (2) Leveraging the known pairs of names and faces from the easy cases using a bootstrapping strategy with additional loss to prevent forgetting and learning new alignments at the same time. We achieve state-of-the-art results for both the augmented Labeled Faces in the Wild dataset and the Celebrity Together dataset. In addition, we believe that our methods can be adapted to other multimodal news understanding tasks.