Interpolation-Based Contrastive Learning for Few-Label Semi-Supervised Learning

Improved Dual Correlation Reduction Network With Affinity Recovery

Deep graph clustering, which aims to reveal the underlying graph structure and divide the nodes into different clusters without human annotations, is a fundamental yet challenging task. However, we observe that the existing methods suffer from the representation collapse problem and tend to encode samples with different classes into the same latent embedding. Consequently, the discriminative capability of nodes is limited, resulting in suboptimal clustering performance. To address this problem, we propose a novel deep graph clustering algorithm termed improved dual correlation reduction network (IDCRN) through improving the discriminative capability of samples. Specifically, by approximating the cross-view feature correlation matrix to an identity matrix, we reduce the redundancy between different dimensions of features, thus improving the discriminative capability of the latent space explicitly. Meanwhile, the cross-view sample correlation matrix is forced to approximate the designed clustering-refined adjacency matrix to guide the learned latent representation to recover the affinity matrix even across views, thus enhancing the discriminative capability of features implicitly. Moreover, we avoid the collapsed representation caused by the oversmoothing issue in graph convolutional networks (GCNs) through an introduced propagation regularization term, enabling IDCRN to capture the long-range information with the shallow network structure. Extensive experimental results on six benchmarks have demonstrated the effectiveness and efficiency of IDCRN compared with the existing state-of-the-art deep graph clustering algorithms. The code of IDCRN is released at IDCRN. Besides, we share a collection of deep graph clustering, including papers, codes, and datasets at ADGC.

Semi-Supervised Learning Under General Causal Models

Semi-supervised learning (SSL) aims to train a machine learning (ML) model using both labeled and unlabeled data. While the unlabeled data have been used in various ways to improve the prediction accuracy, the reason why unlabeled data could help is not fully understood. One interesting and promising direction is to understand SSL from a causal perspective. In light of the independent causal mechanisms (ICM) principle, the unlabeled data can be helpful when the label causes the features but not vice versa. However, the causal relations between the features and labels can be complex in real world applications. In this article, we propose an SSL framework that works with general causal models in which the variables have flexible causal relations. More specifically, we explore the causal graph structures and design corresponding causal generative models which can be learned with the help of unlabeled data. The learned causal generative model can generate synthetic labeled data for training a more accurate predictive model. We verify the effectiveness of our proposed method by empirical studies on both simulated and real data.

One-Step Multi-View Clustering With Diverse Representation

Multi-View clustering has attracted broad attention due to its capacity to utilize consistent and complementary information among views. Although tremendous progress has been made recently, most existing methods undergo high complexity, preventing them from being applied to large-scale tasks. Multi-View clustering via matrix factorization is a representative to address this issue. However, most of them map the data matrices into a fixed dimension, limiting the model's expressiveness. Moreover, a range of methods suffers from a two-step process, i.e., multimodal learning and the subsequent k-means, inevitably causing a suboptimal clustering result. In light of this, we propose a one-step multi-view clustering with diverse representation (OMVCDR) method, which incorporates multi-view learning and k-means into a unified framework. Specifically, we first project original data matrices into various latent spaces to attain comprehensive information and auto-weight them in a self-supervised manner. Then, we directly use the information matrices under diverse dimensions to obtain consensus discrete clustering labels. The unified work of representation learning and clustering boosts the quality of the final results. Furthermore, we develop an efficient optimization algorithm with proven convergence to solve the resultant problem. Comprehensive experiments on various datasets demonstrate the promising clustering performance of our proposed method. The code is publicly available at https://github.com/wanxinhang/OMVCDR.

Toward Balance Deep Semisupervised Clustering

The goal of balanced clustering is partitioning data into distinct groups of equal size. Previous studies have attempted to address this problem by designing balanced regularizers or utilizing conventional clustering methods. However, these methods often rely solely on classic methods, which limits their performance and primarily focuses on low-dimensional data. Although neural networks exhibit effective performance on high-dimensional datasets, they struggle to effectively leverage prior knowledge for clustering with a balanced tendency. To overcome the above limitations, we propose deep semisupervised balanced clustering, which simultaneously learns clustering and generates balance-favorable representations. Our model is based on the autoencoder paradigm incorporating a semisupervised module. Specifically, we introduce a balance-oriented clustering loss and incorporate pairwise constraints into the penalty term as a pluggable module using the Lagrangian multiplier method. Theoretically, we ensure that the proposed model maintains a balanced orientation and provides a comprehensive optimization process. Empirically, we conducted extensive experiments on four datasets to demonstrate significant improvements in clustering performance and balanced measurements. Our code is available at https://github.com/DuannYu/BalancedSemi-TNNLS.

Learning to Predict Gradients for Semi-Supervised Continual Learning

A key challenge for machine intelligence is to learn new visual concepts without forgetting the previously acquired knowledge. Continual learning (CL) is aimed toward addressing this challenge. However, there still exists a gap between CL and human learning. In particular, humans are able to continually learn from the samples associated with known or unknown labels in their daily lives, whereas existing CL and semi-supervised CL (SSCL) methods assume that the training samples are associated with known labels. Specifically, we are interested in two questions: 1) how to utilize unrelated unlabeled data for the SSCL task and 2) how unlabeled data affect learning and catastrophic forgetting in the CL task. To explore these issues, we formulate a new SSCL method, which can be generically applied to existing CL models. Furthermore, we propose a novel gradient learner to learn from labeled data to predict gradients on unlabeled data. In this way, the unlabeled data can fit into the supervised CL framework. We extensively evaluate the proposed method on mainstream CL methods, adversarial CL (ACL), and semi-supervised learning (SSL) tasks. The proposed method achieves state-of-the-art performance on classification accuracy and backward transfer (BWT) in the CL setting while achieving the desired performance on classification accuracy in the SSL setting. This implies that the unlabeled images can enhance the generalizability of CL models on the predictive ability of unseen data and significantly alleviate catastrophic forgetting. The code is available at https://github.com/luoyan407/grad_prediction.git.

Feature-Based Interpolation and Geodesics in the Latent Spaces of Generative Models

Interpolating between points is a problem connected simultaneously with finding geodesics and study of generative models. In the case of geodesics, we search for the curves with the shortest length, while in the case of generative models, we typically apply linear interpolation in the latent space. However, this interpolation uses implicitly the fact that Gaussian is unimodal. Thus, the problem of interpolating in the case when the latent density is non-Gaussian is an open problem. In this article, we present a general and unified approach to interpolation, which simultaneously allows us to search for geodesics and interpolating curves in latent space in the case of arbitrary density. Our results have a strong theoretical background based on the introduced quality measure of an interpolating curve. In particular, we show that maximizing the quality measure of the curve can be equivalently understood as a search of geodesic for a certain redefinition of the Riemannian metric on the space. We provide examples in three important cases. First, we show that our approach can be easily applied to finding geodesics on manifolds. Next, we focus our attention in finding interpolations in pretrained generative models. We show that our model effectively works in the case of arbitrary density. Moreover, we can interpolate in the subset of the space consisting of data possessing a given feature. The last case is focused on finding interpolation in the space of chemical compounds.

Multi-Task Credible Pseudo-Label Learning for Semi-Supervised Crowd Counting

As a widely used semi-supervised learning strategy, self-training generates pseudo-labels to alleviate the labor-intensive and time-consuming annotation problems in crowd counting while boosting the model performance with limited labeled data and massive unlabeled data. However, the noise in the pseudo-labels of the density maps greatly hinders the performance of semi-supervised crowd counting. Although auxiliary tasks, e.g., binary segmentation, are utilized to help improve the feature representation learning ability, they are isolated from the main task, i.e., density map regression and the multi-task relationships are totally ignored. To address the above issues, we develop a multi-task credible pseudo-label learning (MTCP) framework for crowd counting, consisting of three multi-task branches, i.e., density regression as the main task, and binary segmentation and confidence prediction as the auxiliary tasks. Multi-task learning is conducted on the labeled data by sharing the same feature extractor for all three tasks and taking multi-task relations into account. To reduce epistemic uncertainty, the labeled data are further expanded, by trimming the labeled data according to the predicted confidence map for low-confidence regions, which can be regarded as an effective data augmentation strategy. For unlabeled data, compared with the existing works that only use the pseudo-labels of binary segmentation, we generate credible pseudo-labels of density maps directly, which can reduce the noise in pseudo-labels and therefore decrease aleatoric uncertainty. Extensive comparisons on four crowd-counting datasets demonstrate the superiority of our proposed model over the competing methods. The code is available at: https://github.com/ljq2000/MTCP.

Deep Fusion Clustering Network With Reliable Structure Preservation

Deep clustering, which can elegantly exploit data representation to seek a partition of the samples, has attracted intensive attention. Recently, combining auto-encoder (AE) with graph neural networks (GNNs) has accomplished excellent performance by introducing structural information implied among data in clustering tasks. However, we observe that there are some limitations of most existing works: 1) in practical graph datasets, there exist some noisy or inaccurate connections among nodes, which would confuse network learning and cause biased representations, thus leading to unsatisfied clustering performance; 2) lacking dynamic information fusion module to carefully combine and refine the node attributes and the graph structural information to learn more consistent representations; and 3) failing to exploit the two separated views' information for generating a more robust target distribution. To solve these problems, we propose a novel method termed deep fusion clustering network with reliable structure preservation (DFCN-RSP). Specifically, the random walk mechanism is introduced to boost the reliability of the original graph structure by measuring localized structure similarities among nodes. It can simultaneously filter out noisy connections and supplement reliable connections in the original graph. Moreover, we provide a transformer-based graph auto-encoder (TGAE) that can use a self-attention mechanism with the localized structure similarity information to fine-tune the fused topology structure among nodes layer by layer. Furthermore, we provide a dynamic cross-modality fusion strategy to combine the representations learned from both TGAE and AE. Also, we design a triplet self-supervision strategy and a target distribution generation measure to explore the cross-modality information. The experimental results on five public benchmark datasets reflect that DFCN-RSP is more competitive than the state-of-the-art deep clustering algorithms. The corresponding code is available at https://github.com/gongleii/DFCN-RSP.

Semi-Supervised Medical Image Segmentation Guided by Bi-Directional Constrained Dual-Task Consistency

BACKGROUND

Medical image processing tasks represented by multi-object segmentation are of great significance for surgical planning, robot-assisted surgery, and surgical safety. However, the exceptionally low contrast among tissues and limited available annotated data makes developing an automatic segmentation algorithm for pelvic CT challenging.

METHODS

A bi-direction constrained dual-task consistency model named PICT is proposed to improve segmentation quality by leveraging free unlabeled data. First, to learn more unmarked data features, it encourages the model prediction of the interpolated image to be consistent with the interpolation of the model prediction at the pixel, model, and data levels. Moreover, to constrain the error prediction of interpolation interference, PICT designs an auxiliary pseudo-supervision task that focuses on the underlying information of non-interpolation data. Finally, an effective loss algorithm for both consistency tasks is designed to ensure the complementary manner and produce more reliable predictions.

RESULTS

Quantitative experiments show that the proposed PICT achieves 87.18%, 96.42%, and 79.41% mean DSC score on ACDC, CTPelvic1k, and the individual Multi-tissue Pelvis dataset with gains of around 0.8%, 0.5%, and 1% compared to the state-of-the-art semi-supervised method. Compared to the baseline supervised method, the PICT brings over 3-9% improvements.

CONCLUSIONS

The developed PICT model can effectively leverage unlabeled data to improve segmentation quality of low contrast medical images. The segmentation result could improve the precision of surgical path planning and provide input for robot-assisted surgery.