Learning Discriminative Representations Research Articles

Accurate pneumoconiosis staging via deep texture encoding and discriminative representation learning.

Accurate pneumoconiosis staging is key to early intervention and treatment planning for pneumoconiosis patients. The staging process relies on assessing the profusion level of small opacities, which are dispersed throughout the entire lung field and manifest as fine textures. While conventional convolutional neural networks (CNNs) have achieved significant success in tasks such as image classification and object recognition, they are less effective for classifying fine-grained medical images due to the need for global, orderless feature representation. This limitation often results in inaccurate staging outcomes for pneumoconiosis. In this study, we propose a deep texture encoding scheme with a suppression strategy designed to capture the global, orderless characteristics of pneumoconiosis lesions while suppressing prominent regions such as the ribs and clavicles within the lung field. To further enhance staging accuracy, we incorporate an ordinal label distribution to capture the ordinal information among profusion levels of opacities. Additionally, we employ supervised contrastive learning to develop a more discriminative feature space for downstream classification tasks. Finally, in accordance with standard practices, we evaluate the profusion levels of opacities in each subregion of the lung, rather than relying on the entire chest X-ray image. Experimental results on the pneumoconiosis dataset demonstrate the superior performance of the proposed method confirming its effectiveness for accurate pneumoconiosis staging.

Zero-shot visual grounding via coarse-to-fine representation learning

Visual grounding (VG) locates target objects in visual scenes by understanding given natural language queries. Current methods for VG mainly focus on grounding referring expressions or noun phrases covered in the labeled training samples. Despite their grounding prowess, these approaches struggle in grounding novel query–image pairs excluded from the training data. This shortage is usually caused by the deficiency of discriminative representation learning both in images and queries. To address these issues, we propose a one-stage coarse-to-fine framework for zero-shot VG to ground novel query-image samples. Specifically, in the coarse stage, we mine the global context information in the visual features and query embeddings by employing a multi-head self-attention block, strengthening the intra-modality relations in the visual and textual features. In the fine stage, we first learn the query-aware visual representations based on the acquired global context information via a multi-modal relation-enhanced transformer block, which explores the informative information by modeling the cross-modal interaction among visual and textual domains. We further excavate target-oriented discriminative representations from the acquired query-aware visual representations by a noun phrase-guided multi-modal interaction network, which augments the interaction between target-related phrases and the obtained query-aware visual representations to enhance the distinction of target regions, enhancing the subsequent referred target grounding and generalizing. In order to validate the proposed approach, we implement extensive experiments and ablation studies on public benchmark datasets, including RefCOCO, RefCOCO+, RefCOCOg, Flickr30K Entity, Flickr-Split-0 and Flickr-Split-1. Experimental results demonstrate that our approach substantially improves the grounding accuracy and achieves new state-of-the-art performance under single-stage training and testing.

Progressive Neighbor-masked Contrastive Learning for Fusion-style Deep Multi-view Clustering

Fusion-style Deep Multi-view Clustering (FDMC) can efficiently integrate comprehensive feature information from latent embeddings of multiple views and has drawn much attention recently. However, existing FDMC methods suffer from the interference of view-specific information for fusion representation, affecting the learning of discriminative cluster structure. In this paper, we propose a new framework of Progressive Neighbor-masked Contrastive Learning for FDMC (PNCL-FDMC) to tackle the aforementioned issues. Specifically, by using neighbor-masked contrastive learning, PNCL-FDMC can explicitly maintain the local structure during the embedding aggregation, which is beneficial to the common semantics enhancement on the fusion view. Based on the consistent aggregation, the fusion view is further enhanced by diversity-aware cluster structure enhancement. In this process, the enhanced cluster assignments and cluster discrepancies are employed to guide the weighted neighbor-masked contrastive alignment of semantic structure between individual views and the fusion view. Extensive experiments validate the effectiveness of the proposed framework, revealing its ability in discriminative representation learning and improving clustering performance.

Mixed Graph Contrastive Network for Semi-supervised Node Classification

Graph Neural Networks (GNNs) have achieved promising performance in semi-supervised node classification in recent years. However, the problem of insufficient supervision, together with representation collapse, largely limits the performance of the GNNs in this field. To alleviate the collapse of node representations in semi-supervised scenario, we propose a novel graph contrastive learning method, termed M ixed G raph C ontrastive N etwork (MGCN). In our method, we improve the discriminative capability of the latent embeddings by an interpolation-based augmentation strategy and a correlation reduction mechanism. Specifically, we first conduct the interpolation-based augmentation in the latent space and then force the prediction model to change linearly between samples. Second, we enable the learned network to tell apart samples across two interpolation-perturbed views through forcing the correlation matrix across views to approximate an identity matrix. By combining the two settings, we extract rich supervision information from both the abundant unlabeled nodes and the rare yet valuable labeled nodes for discriminative representation learning. Extensive experimental results on six datasets demonstrate the effectiveness and the generality of MGCN compared to the existing state-of-the-art methods. The code of MGCN is available at https://github.com/xihongyang1999/MGCN on Github.

Edge‐guided representation learning for underwater object detection

AbstractUnderwater object detection (UOD) is crucial for marine economic development, environmental protection, and the planet's sustainable development. The main challenges of this task arise from low‐contrast, small objects, and mimicry of aquatic organisms. The key to addressing these challenges is to focus the model on obtaining more discriminative information. The authors observe that the edges of underwater objects are highly unique and can be distinguished from low‐contrast or mimicry environments based on their edges. Motivated by this observation, an Edge‐guided Representation Learning Network, termed ERL‐Net is proposed, that aims to achieve discriminative representation learning and aggregation under the guidance of edge cues. Firstly, an edge‐guided attention module is introduced to model the explicit boundary information, which generates more discriminative features. Secondly, a hierarchical feature aggregation module is proposed to aggregate the multi‐scale discriminative features by regrouping them into three levels, effectively aggregating global and local information for locating and recognising underwater objects. Finally, a wide and asymmetric receptive field block is proposed to enable features to have a wider receptive field, allowing the model to focus on smaller object information. Comprehensive experiments on three challenging underwater datasets show that our method achieves superior performance on the UOD task.

TFPred: Learning discriminative representations from unlabeled data for few-label rotating machinery fault diagnosis

Recent advances in intelligent rotating machinery fault diagnosis have been enabled by the availability of massive labeled training data. However, in practical industrial applications, it is often challenging and costly to annotate a large amount of data. To address the few-label fault diagnosis problem, a time–frequency prediction (TFPred) self-supervised learning framework is proposed to extract latent fault representations from unlabeled fault data. Specifically, the TFPred framework consists of a time encoder and a frequency encoder, with the frequency encoder to predict the low-dimensional representations of time domain signals generated by the time encoder with randomly augmented data. Subsequently, the pre-trained network is hooked with a classification head and fine-tuned with limited labeled data. Finally, the proposed framework is evaluated on a run-to-failure bearing dataset and a hardware-in-the-loop high-speed train simulation platform. The experiments demonstrate that the self-supervised learning framework TFPred achieved competitive performance with only 1% and 5% labeled data. Code is available at https://github.com/Xiaohan-Chen/TFPred.

Time Interval-Enhanced Graph Neural Network for Shared-Account Cross-Domain Sequential Recommendation.

Shared-account cross-domain sequential recommendation (SCSR) task aims to recommend the next item via leveraging the mixed user behaviors in multiple domains. It is gaining immense research attention as more and more users tend to sign up on different platforms and share accounts with others to access domain-specific services. Existing works on SCSR mainly rely on mining sequential patterns via recurrent neural network (RNN)-based models, which suffer from the following limitations: 1) RNN-based methods overwhelmingly target discovering sequential dependencies in single-user behaviors and they are not expressive enough to capture the relationships among multiple entities in SCSR; 2) all existing methods bridge two domains via knowledge transfer in the latent space and ignore the explicit cross-domain graph structure; and 3) none existing studies consider the time interval information among items, which is essential in the sequential recommendation for characterizing different items and learning discriminative representations for them. In this work, we propose a new graph-based solution, namely, time interval-enhanced domain-aware graph convolutional network (TiDA-GCN), to address the above challenges. Specifically, we first link users and items in each domain as a graph. Then, we devise a domain-aware graph convolution network to learn user-specific node representations. To fully account for users' domain-specific preferences on items, two effective attention mechanisms are further developed to selectively guide the message-passing process. Moreover, to further enhance item- and account-level representation learning, we incorporate the time interval into the message passing and design an account-aware self-attention module for learning items' interactive characteristics. Experiments demonstrate the superiority of our proposed method from various aspects.

Duplex-Hierarchy Representation Learning for Remote Sensing Image Classification.

Remote sensing image classification (RSIC) is designed to assign specific semantic labels to aerial images, which is significant and fundamental in many applications. In recent years, substantial work has been conducted on RSIC with the help of deep learning models. Even though these models have greatly enhanced the performance of RSIC, the issues of diversity in the same class and similarity between different classes in remote sensing images remain huge challenges for RSIC. To solve these problems, a duplex-hierarchy representation learning (DHRL) method is proposed. The proposed DHRL method aims to explore duplex-hierarchy spaces, including a common space and a label space, to learn discriminative representations for RSIC. The proposed DHRL method consists of three main steps: First, paired images are fed to a pretrained ResNet network for extracting the corresponding features. Second, the extracted features are further explored and mapped into a common space for reducing the intra-class scatter and enlarging the inter-class separation. Third, the obtained representations are used to predict the categories of the input images, and the discrimination loss in the label space is minimized to further promote the learning of discriminative representations. Meanwhile, a confusion score is computed and added to the classification loss for guiding the discriminative representation learning via backpropagation. The comprehensive experimental results show that the proposed method is superior to the existing state-of-the-art methods on two challenging remote sensing image scene datasets, demonstrating that the proposed method is significantly effective.

A multi-modal feature fusion classification model based on distance matching and discriminative representation learning for differentiation of high-grade glioma from solitary brain metastasis

To explore the performance of a new multimodal feature fusion classification model based on distance matching and discriminative representation learning for differentiating high-grade glioma (HGG) from solitary brain metastasis (SBM). We collected multi-parametric magnetic resonance imaging (MRI) data from 61 patients with HGG and 60 with SBM, and delineated regions of interest (ROI) on T1WI, T2WI, T2-weighted fluid attenuated inversion recovery (T2_FLAIR) and post-contrast enhancement T1WI (CE_T1WI) images. The radiomics features were extracted from each sequence using Pyradiomics and fused using a multimodal feature fusion classification model based on distance matching and discriminative representation learning to obtain a classification model. The discriminative performance of the classification model for differentiating HGG from SBM was evaluated using five-fold cross-validation with metrics of specificity, sensitivity, accuracy, and the area under the ROC curve (AUC) and quantitatively compared with other feature fusion models. Visual experiments were conducted to examine the fused features obtained by the proposed model to validate its feasibility and effectiveness. The five-fold cross-validation results showed that the proposed multimodal feature fusion classification model had a specificity of 0.871, a sensitivity of 0.817, an accuracy of 0.843, and an AUC of 0.930 for distinguishing HGG from SBM. This feature fusion method exhibited excellent discriminative performance in the visual experiments. The proposed multimodal feature fusion classification model has an excellent ability for differentiating HGG from SBM with significant advantages over other feature fusion classification models in discrimination and classification tasks between HGG and SBM.

SpatiAlign: an unsupervised contrastive learning model for data integration of spatially resolved transcriptomics.

Integrative analysis of spatially resolved transcriptomics datasets empowers a deeper understanding of complex biological systems. However, integrating multiple tissue sections presents challenges for batch effect removal, particularly when the sections are measured by various technologies or collected at different times. We propose spatiAlign, an unsupervised contrastive learning model that employs the expression of all measured genes and the spatial location of cells, to integrate multiple tissue sections. It enables the joint downstream analysis of multiple datasets not only in low-dimensional embeddings but also in the reconstructed full expression space. In benchmarking analysis, spatiAlign outperforms state-of-the-art methods in learning joint and discriminative representations for tissue sections, each potentially characterized by complex batch effects or distinct biological characteristics. Furthermore, we demonstrate the benefits of spatiAlign for the integrative analysis of time-series brain sections, including spatial clustering, differential expression analysis, and particularly trajectory inference that requires a corrected gene expression matrix.

A Novel Hierarchical Cross-Stream Aggregation Neural Network for Semantic Segmentation of 3-D Dental Surface Models.

Accurate teeth delineation on 3-D dental models is essential for individualized orthodontic treatment planning. Pioneering works like PointNet suggest a promising direction to conduct efficient and accurate 3-D dental model analyses in end-to-end learnable fashions. Recent studies further imply that multistream architectures to concurrently learn geometric representations from different inputs/views (e.g., coordinates and normals) are beneficial for segmenting teeth with varying conditions. However, such multistream networks typically adopt simple late-fusion strategies to combine features captured from raw inputs that encode complementary but fundamentally different geometric information, potentially hampering their accuracy in end-to-end semantic segmentation. This article presents a hierarchical cross-stream aggregation (HiCA) network to learn more discriminative point/cell-wise representations from multiview inputs for fine-grained 3-D semantic segmentation. Specifically, based upon our multistream backbone with input-tailored feature extractors, we first design a contextual cross-steam aggregation (CA) module conditioned on interstream consistency to boost each view's contextual representation learning jointly. Then, before the late fusion of different streams' outputs for segmentation, we further deploy a discriminative cross-stream aggregation (DA) module to concurrently update all views' discriminative representation learning by leveraging a specific graph attention strategy induced by multiview prototype learning. On both public and in-house datasets of real-patient dental models, our method significantly outperformed state-of-the-art (SOTA) deep learning methods for teeth semantic segmentation. In addition, extended experimental results suggest the applicability of HiCA to other general 3-D shape segmentation tasks. The code is available at https://github.com/ladderlab-xjtu/HiCA.

Learning Discriminative Representations From Cross-Scale Features for Camouflaged Object Detection

Learning Discriminative Representations From Cross-Scale Features for Camouflaged Object Detection

Generalizable and Discriminative Representations for Adversarially Robust Few-Shot Learning.

Few-shot image classification (FSIC) is beneficial for a variety of real-world scenarios, aiming to construct a recognition system with limited training data. In this article, we extend the original FSIC task by incorporating defense against malicious adversarial examples. This can be an arduous challenge because numerous deep learning-based approaches remain susceptible to adversarial examples, even when trained with ample amounts of data. Previous studies on this problem have predominantly concentrated on the meta-learning framework, which involves sampling numerous few-shot tasks during the training stage. In contrast, we propose a straightforward but effective baseline via learning robust and discriminative representations without tedious meta-task sampling, which can further be generalized to unforeseen adversarial FSIC tasks. Specifically, we introduce an adversarial-aware (AA) mechanism that exploits feature-level distinctions between the legitimate and the adversarial domains to provide supplementary supervision. Moreover, we design a novel adversarial reweighting training strategy to ameliorate the imbalance among adversarial examples. To further enhance the adversarial robustness without compromising discriminative features, we propose the cyclic feature purifier during the postprocessing projection, which can reduce the interference of unforeseen adversarial examples. Furthermore, our method can obtain robust feature embeddings that maintain superior transferability, even when facing cross-domain adversarial examples. Extensive experiments and systematic analyses demonstrate that our method achieves state-of-the-art robustness as well as natural performance among adversarially robust FSIC algorithms on three standard benchmarks by a substantial margin.

Weighted Graph-Structured Semantics Constraint Network for Cross-Modal Retrieval

Cross-modal retrieval aims to retrieve relevant content of different modalities by giving a query of another modality. The biggest difficulty is how to bridge the heterogeneous gap between different modalities. The commonly-used methods tend to focus on exploiting individual image-text pair and mining the relations of cross-modality data thereof, but ignore the role of multi-sample correlation. Moreover, more global, structural inter-pair knowledge contained by the training dataset will be under-used. To fully exploit graph-structured semantics and mine the semantic information in the dataset for learning discriminative representations, we propose Weighted Graph-structured Semantics Constraint Network (WGSCN), a unified, graph-based, semantic-constrained learning framework,in which GCN is used to mine comprehensive relation information from cross modality data. Our main inspiration is to design a novel two-branch GCN-based Cross-modal Semantic Encoding (GCSE) module to produce semantic embeddings with the both modality-specific and modality-shared correlation. Moreover, a GAN-based dual learning approach is used to further improve the discriminability and model the joint distribution across different modalities. Our proposed GDL uses semantic embeddings as supervisory signal to make the common representation semantically discriminative while adversarial learning and dual learning are used to make the common representation modality-invariant. Through comparative experiments on five commonly used cross-modal datasets, we have shown the superior retrieval accuracy of our WGSCN.

Federated Discriminative Representation Learning for Image Classification.

Acquiring big-size datasets to raise the performance of deep models has become one of the most critical problems in representation learning (RL) techniques, which is the core potential of the emerging paradigm of federated learning (FL). However, most current FL models concentrate on seeking an identical model for isolated clients and thus fail to make full use of the data specificity between clients. To enhance the classification performance of each client, this study introduces the FDRL, a federated discriminative RL model, by partitioning the data features of each client into a global subspace and a local subspace. More specifically, FDRL learns the global representation for federated communication between those isolated clients, which is to capture common features from all protected datasets via model sharing, and local representations for personalization in each client, which is to preserve specific features of clients via model differentiating. Toward this goal, FDRL in each client trains a shared submodel for federated communication and, meanwhile, a not-shared submodel for locality preservation, in which the two models partition client-feature space by maximizing their differences, followed by a linear model fed with combined features for image classification. The proposed model is implemented with neural networks and optimized in an iterative manner between the server of computing the global model and the clients of learning the local classifiers. Thanks to the powerful capability of local feature preservation, FDRL leads to more discriminative data representations than the compared FL models. Experimental results on public datasets demonstrate that our FDRL benefits from the subspace partition and achieves better performance on federated image classification than the state-of-the-art FL models.

Self-Supervised Contrastive Representation Learning for Semi-Supervised Time-Series Classification.

Learning time-series representations when only unlabeled data or few labeled samples are available can be a challenging task. Recently, contrastive self-supervised learning has shown great improvement in extracting useful representations from unlabeled data via contrasting different augmented views of data. In this work, we propose a novel Time-Series representation learning framework via Temporal and Contextual Contrasting (TS-TCC) that learns representations from unlabeled data with contrastive learning. Specifically, we propose time-series-specific weak and strong augmentations and use their views to learn robust temporal relations in the proposed temporal contrasting module, besides learning discriminative representations by our proposed contextual contrasting module. Additionally, we conduct a systematic study of time-series data augmentation selection, which is a key part of contrastive learning. We also extend TS-TCC to the semi-supervised learning settings and propose a Class-Aware TS-TCC (CA-TCC) that benefits from the available few labeled data to further improve representations learned by TS-TCC. Specifically, we leverage the robust pseudo labels produced by TS-TCC to realize a class-aware contrastive loss. Extensive experiments show that the linear evaluation of the features learned by our proposed framework performs comparably with the fully supervised training. Additionally, our framework shows high efficiency in few labeled data and transfer learning scenarios. The code is publicly available at https://github.com/emadeldeen24/CA-TCC.

Joint discriminative representation learning for end-to-end person search

Person search simultaneously detects and retrieves a query person from uncropped scene images. Existing methods are either two-step or end-to-end. The former employs two standalone models for the two sub-tasks, while the latter conducts person search with a unified model. Despite encouraging progress, most existing end-to-end methods focus on balancing the model between detection and retrieval sub-tasks, while ignoring to enhance the learned representation for retrieval, which leads to inferior accuracy to two-step approaches. To that end, we propose a novel hierarchical framework that jointly optimizes instance-aware and part-aware embedding to enable discriminative representation learning. Specifically, we develop a region-of-interest cosegment (ROICoseg) module that captures part-aware information without requiring extra annotations to enable fine-grained discriminative representation. On top of that, a Contextual Instance Batch Sampling (CIBS) method is introduced to effectively employ contextual information for constructing training batches, thus facilitating effective instance-aware representation learning. We further introduce the first cross-door person search dataset (CDPS) that retrieves a target person in outdoor cameras with an indoor captured image or vice versa. Extensive experiments show that our proposed model achieves competitive performance on CUHK-SYSU and outperforms state-of-the-art end-to-end methods on the more challenging PRW and CDPS.11Code and dataset are available at https://github.com/PatrickZad/CDPS.

Contrastive sentence representation learning with adaptive false negative cancellation

Contrastive sentence representation learning has made great progress thanks to a range of text augmentation strategies and hard negative sampling techniques. However, most studies directly employ in-batch samples as negative samples, ignoring the semantic relationship between negative samples and anchors, which may lead to negative sampling bias. To address this issue, we propose similarity and relative-similarity strategies for identifying potential false negatives. Moreover, we introduce adaptive false negative elimination and attraction methods to mitigate their adverse effects. Our proposed approaches can also be considered semi-supervised contrastive learning, as the identified false negatives can be viewed as either negative or positive samples for contrastive learning in adaptive false negative elimination and attraction methods. By fusing information from positive and negative pairs, contrastive learning learns rich and discriminative representations that capture the intrinsic characteristics of the sentence. Experimental results indicate that our proposed strategies and methods can bring further significant performance improvements. Specifically, the combination of similarity strategy and adaptive false negative elimination method achieves the best results, yielding an average performance gain of 2.1% compared to SimCSE in semantic textual similarity (STS) tasks. Furthermore, our approach is generalizable and can be applied to different text data augmentation strategies and certain existing contrastive sentence representation learning models. Our experimental code and data are publicly available at the link: https://github.com/Linda230/AFNC.

GRAB-Net: Graph-based Boundary-aware Network for Medical Point Cloud Segmentation.

Point cloud segmentation is fundamental in many medical applications, such as aneurysm clipping and orthodontic planning. Recent methods mainly focus on designing powerful local feature extractors and generally overlook the segmentation around the boundaries between objects, which is extremely harmful to the clinical practice and degenerates the overall segmentation performance. To remedy this problem, we propose a GRAph-based Boundary-aware Network (GRAB-Net) with three paradigms, Graph-based Boundary-perception Module (GBM), Outer-boundary Context-assignment Module (OCM), and Inner-boundary Feature-rectification Module (IFM), for medical point cloud segmentation. Aiming to improve the segmentation performance around boundaries, GBM is designed to detect boundaries and interchange complementary information inside semantic and boundary features in the graph domain, where semantics-boundary correlations are modelled globally and informative clues are exchanged by graph reasoning. Furthermore, to reduce the context confusion that degenerates the segmentation performance outside the boundaries, OCM is proposed to construct the contextual graph, where dissimilar contexts are assigned to points of different categories guided by geometrical landmarks. In addition, we advance IFM to distinguish ambiguous features inside boundaries in a contrastive manner, where boundary-aware contrast strategies are proposed to facilitate the discriminative representation learning. Extensive experiments on two public datasets, IntrA and 3DTeethSeg, demonstrate the superiority of our method over state-of-the-art methods.

Multi-Constraint Latent Representation Learning for Prognosis Analysis Using Multi-Modal Data.

The Cox proportional hazard model has been widely applied to cancer prognosis prediction. Nowadays, multi-modal data, such as histopathological images and gene data, have advanced this field by providing histologic phenotype and genotype information. However, how to efficiently fuse and select the complementary information of high-dimensional multi-modal data remains challenging for Cox model, as it generally does not equip with feature fusion/selection mechanism. Many previous studies typically perform feature fusion/selection in the original feature space before Cox modeling. Alternatively, learning a latent shared feature space that is tailored for Cox model and simultaneously keeps sparsity is desirable. In addition, existing Cox-based models commonly pay little attention to the actual length of the observed time that may help to boost the model's performance. In this article, we propose a novel Cox-driven multi-constraint latent representation learning framework for prognosis analysis with multi-modal data. Specifically, for efficient feature fusion, a multi-modal latent space is learned via a bi-mapping approach under ranking and regression constraints. The ranking constraint utilizes the log-partial likelihood of Cox model to induce learning discriminative representations in a task-oriented manner. Meanwhile, the representations also benefit from regression constraint, which imposes the supervision of specific survival time on representation learning. To improve generalization and alleviate overfitting, we further introduce similarity and sparsity constraints to encourage extra consistency and sparseness. Extensive experiments on three datasets acquired from The Cancer Genome Atlas (TCGA) demonstrate that the proposed method is superior to state-of-the-art Cox-based models.