Adapting person re-identification (reID) models to new target environments remains a challenging problem that is typically addressed using unsupervised domain adaptation (UDA) methods. Recent works show that when labeled data originates from several distinct sources (e.g., datasets and cameras), considering each source separately and applying multi-source domain adaptation (MSDA) typically yields higher accuracy and robustness compared to blending the sources and performing conventional UDA. However, state-of-the-art MSDA methods learn domain-specific backbone models or require access to source domain data during adaptation, resulting in significant growth in training parameters and computational cost. In this paper, a Source-free Adaptive Gated Experts (SAGE-reID) method is introduced for person reID. Our SAGE-reID is a cost-effective, source-free MSDA method that first trains individual source-specific low-rank adapters (LoRA) through source-free UDA. Then, a lightweight gating network is introduced and trained to dynamically assign optimal merging weights for fusion of LoRA experts, enabling effective cross-domain knowledge transfer. While the number of backbone parameters remains constant across source domains, LoRA experts scale linearly but remain negligible in size (ď 2% per source), reducing both the memory consumption and risk of overfitting. Extensive experiments conducted on three challenging benchmarks – Market-1501, DukeMTMC-reID, and MSMT17 – indicate that SAGE-reID can outperform state-of-the-art methods while remaining computationally efficient. Our code is available: https://github.com/nehdiii/SAGE-reID

Graph Self-Supervised Learning (GSSL) has emerged as a powerful paradigm for generating high-quality representations for graph-structured data. While multi-scale graph contrastive learning has received increasing attention, many existing methods still predominantly focus on a single graph abstraction level. To address this limitation, we propose a unified contrastive framework that can target node-level, proximity-level, cluster-level, and graph-level information and integrate them through a linear combination of similarity scores on positive pairs and dissimilarity scores (i.e., similarity scores on negative pairs). Furthermore, current approaches typically assign uniform penalty strengths to all examples, which reduces optimization flexibility and leads to ambiguous convergence status. To overcome this, we introduce a novel parameter-free fine-grained self-weighting mechanism that adaptively assigns weights to individual similarity and dissimilarity scores. The proposed mechanism emphasizes the scores that deviate significantly from their target values. Our approach not only enhances optimization flexibility but also eliminates the computational overhead of hyperparameter tuning in conventional multi-task GSSL methods. Comprehensive experiments on real-world datasets show that our methods consistently outperform state-of-the-art approaches across downstream tasks, including classification, clustering, and link prediction, in both single-level and multi-level scenarios.

Subspace clustering methods face inherent scal-ability limits due to the$O(n^{3})$cost (with$n$denoting the number of data samples) of constructing full$n\times n$affinities and performing spectral decomposition. While deep learning-based approaches improve feature extraction, they maintain this computational bottleneck through exhaustive pairwise similarity computations. We propose SDSNet (Scalable Deep Subspace Network), a deep subspace clustering framework that achieves$\mathrm{O}(n)$complexity through (1) landmark-based approximation, avoiding full affinity matrices, (2) joint optimization of auto-encoder reconstruction with self-expression objectives, and (3) direct spectral clustering on factorized representations. The framework combines convolutional auto-encoders with subspace-preserving constraints. Experimental results demonstrate that SDSNet achieves comparable clustering quality to state-of-the-art methods with significantly improved computational efficiency.

Adapting Vision-Language Models (VLMs) to new domains with few labeled samples remains a significant challenge due to severe overfitting and computational constraints. State-of-the-art solutions, such as low-rank reparameterization, mitigate these issues but often struggle with generalization and require extensive hyperparameter tuning. In this paper, a novel Sparse Optimization (SO) framework is proposed. Unlike low-rank approaches that typically constrain updates to a fixed subspace, our SO method leverages high sparsity to dynamically adjust very few parameters. We introduce two key paradigms. First, we advocate for \textit{local sparsity and global density}, which updates a minimal subset of parameters per iteration while maintaining overall model expressiveness. As a second paradigm, we advocate for \textit{local randomness and global importance}, which sparsifies the gradient using random selection while pruning the first moment based on importance. This combination significantly mitigates overfitting and ensures stable adaptation in low-data regimes. Extensive experiments on 11 diverse datasets show that SO achieves state-of-the-art few-shot adaptation performance while reducing memory overhead.

High-dimensional multiplex graphs are characterized by their high number of complementary and divergent dimensions. The existence of multiple hierarchical latent relations between the graph dimensions poses significant challenges to embedding methods. In particular, the geometric distortions that might occur in the representational space have been overlooked in the literature. This work studies the problem of high-dimensional multiplex graph embedding from a geometric perspective. We find that the node representations reside on highly curved manifolds, thus rendering their exploitation more challenging for downstream tasks. Moreover, our study reveals that increasing the number of graph dimensions can cause further distortions to the highly curved manifolds. To address this problem, we propose a novel multiplex graph embedding method that harnesses hierarchical dimension embedding and Hyperbolic Graph Neural Networks. The proposed approach hierarchically extracts hyperbolic node representations that reside on Riemannian manifolds while gradually learning fewer and more expressive latent dimensions of the multiplex graph. Experimental results on real-world high-dimensional multiplex graphs show that the synergy between hierarchical and hyperbolic embeddings incurs much fewer geometric distortions and brings notable improvements over state-of-the-art approaches on downstream tasks.

We investigate the problem of multiplex graph embedding, that is, graphs in which nodes interact through multiple types of relations (dimensions). In recent years, several methods have been developed to address this problem. However, the need for more effective and specialized approaches grows with the production of graph data with diverse characteristics. In particular, real-world multiplex graphs may exhibit a high number of dimensions, making it difficult to construct a single consensus representation. Furthermore, important information can be hidden in complex latent structures scattered in multiple dimensions. To address these issues, we propose HMGE, a novel embedding method based on hierarchical aggregation for high-dimensional multiplex graphs. Hierarchical aggregation consists in learning a hierarchical combination of the graph dimensions and refining the embeddings at each hierarchy level. Non-linear combinations are computed from previous ones, thus uncovering complex information and latent structures hidden in the multiplex graph dimensions. Moreover, we leverage mutual information maximization between local patches and global summaries to train the model without supervision. This allows to captures globally relevant information present in diverse locations of the graph. Detailed experiments on synthetic and real-world data illustrate the suitability of our approach on downstream supervised tasks, including link prediction and node classification.

The most recent approaches for clustering single-cell RNA-sequencing data rely on deep auto-encoders. However, three major challenges remain unaddressed. First, current models overlook the impact of the cumulative errors induced by the pseudo-supervised embedding clustering task (Feature Randomness). Second, existing methods neglect the effect of the strong competition between embedding clustering and reconstruction (Feature Drift). Third, the previous deep clustering models regularly fail to consider the topological information of the latent data, even though the local and global latent configurations can bring complementary views to the clustering task. To address these challenges, we propose a novel approach that explores the interaction between local and global latent configurations to progressively adjust the reconstruction and embedding clustering tasks. We elaborate a topological and probabilistic filter to mitigate Feature Randomness and a cell-cell graph structure and content correction mechanism to counteract Feature Drift. The Zero-Inflated Negative Binomial model is also integrated to capture the characteristics of gene expression profiles. We conduct detailed experiments on real-world datasets from multiple representative genome sequencing platforms. Our approach outperforms the state-of-the-art clustering methods in various evaluation metrics.

Deep clustering models are trained based on self-supervision and pseudo-supervision. However, applying these techniques can cause Feature Randomness and Feature Drift. On one hand, Feature Randomness takes place when a considerable portion of the pseudo-labels do not match the true ones. On the other hand, Feature Drift takes place when there is a strong con-flict between the self-supervision and pseudo-supervision tasks. We propose ADEC (Adversarial Deep Embedded Clustering) a novel autoencoder-based clustering model, which relies on a discriminator network to reduce random features while avoiding the drifting effect. Experimental results validate that our model alleviates these problems and outperforms existing methods.

Recent graph clustering methods have resorted to Graph Auto-Encoders (GAEs). However, two important issues have been overlooked. First, the accumulative error, inflicted by learning from noisy clustering assignments, degrades the model’s effectiveness. This problem is called Feature Randomness (FR). Second, reconstructing the adjacency matrix sets the model to learn irrelevant similarities for the clustering task. This problem is called Feature Drift (FD). To address these issues, we first propose a sampling operator that triggers a protection mechanism against FR. second, we propose an operator Υ that triggers a correction mechanism against FD by transforming the reconstructed graph. Experimental results validate that our operators alleviate these problems and bring significant clustering improvement.

The deep clustering paradigm has shown great potential for discovering complex patterns that can reveal cell heterogeneity in single-cell RNA sequencing data. This paradigm involves two training phases: pretraining based on a pretext task and fine-tuning using pseudo-labels. Although current models yield promising results, they overlook the geometric distortions that regularly occur during the training process. More precisely, the transition between the two phases results in a coarse flattening of the latent structures, which can deteriorate the clustering performance. In this context, existing methods perform euclidean-based embedding clustering without ensuring the flatness and convexity of the latent manifolds. To address this problem, we incorporate two mechanisms. First, we introduce an overclustering loss to flatten the local curves. Second, we propose an adversarial mechanism to adjust the global geometric configuration. The second mechanism gradually transforms the latent structures into convex ones. Empirical results on a variety of gene expression datasets show that our model outperforms state-of-the-art methods.

Variational Graph Auto-Encoders (VGAEs) have achieved promising performance in several applications. Some recent models incorporate the clustering inductive bias by imposing non-Gaussian prior distributions. However, the regularization term is practically insufficient to learn the clustering structures due to the mismatch between the target and the learned distributions. Thus, we formulate a new variational lower bound that incorporates an explicit clustering objective function. The introduction of a clustering objective leads to two problems. First, the latent information destroyed by the clustering process is critical for generating the between-cluster edges. Second, the noisy and sparse input graph does not benefit from the information learned during the clustering process. To address the first problem, we identify a new term overlooked by existing Evidence Lower BOunds (ELBOs). This term accounts for the difference between the variational posterior used for the clustering task and the variational posterior associated with the generation task. Furthermore, we find that the new term increases resistance to posterior collapse. Theoretically, we demonstrate that our lower bound is a tighter approximation of the log-likelihood function. To address the second problem, we propose a graph update algorithm that reduces the over-segmentation and under-segmentation problems. We conduct several experiments to validate the merits of our approach. Our results show that the proposed method considerably improves the clustering quality compared to state-of-the-art VGAE models.

Most recent graph clustering methods have resorted to Graph Auto-Encoders (GAEs) to perform joint clustering and embedding learning. However, two critical issues have been overlooked. First, the accumulative error, inflicted by learning from noisy clustering assignments, degrades the effectiveness of the clustering model. This problem is called Feature Randomness. Second, reconstructing the adjacency matrix sets the model to learn irrelevant similarities for the clustering task. This problem is called Feature Drift. Furthermore, the theoretical relation between the aforementioned problems has not yet been investigated. We study these issues from two aspects: (1) there is a trade-off between Feature Randomness and Feature Drift when clustering and reconstruction are performed at the same level, and (2) the problem of Feature Drift is more pronounced for GAE models, compared with vanilla auto-encoder models. Thus, we reformulate the GAE-based clustering methodology. Our solution is two-fold. First, we propose a sampling operator$\Xi$that triggers a protection mechanism against Feature Randomness. Second, we propose an operator$\Upsilon$that triggers a correction mechanism against Feature Drift by gradually transforming the reconstructed graph into a clustering-oriented one. As principal advantages, our solution grants a considerable improvement in clustering effectiveness and can be easily tailored to GAE models.

Most recent graph clustering methods rely on pretraining graph auto-encoders using self-supervision techniques (pretext task) and finetuning based on pseudo-supervision (main task). However, the transition from self-supervision to pseudo-supervision has never been studied from a geometric perspective. Herein, we establish the first systematic exploration of the latent manifolds' geometry under the deep clustering paradigm; we study the evolution of their intrinsic dimension and linear intrinsic dimension. We find that the embedded manifolds undergo coarse geometric transformations under the transition regime: from curved low-dimensional to flattened higher-dimensional. Moreover, we find that this inappropriate flattening leads to clustering deterioration by twisting the curved structures. To address this problem, which we call Feature Twist, we propose a variational graph auto-encoder that can smooth the local curves before gradually flattening the global structures. Our results show a notable improvement over multiple state-of-the-art approaches by escaping Feature Twist.

To overcome the absence of concrete supervisory signals, deep clustering models construct their own labels based on self-supervision and pseudo-supervision. However, applying these techniques can cause Feature Randomness and Feature Drift. In this paper, we formally characterize these two new concepts. On one hand, Feature Randomness takes place when a considerable portion of the pseudo-labels is deemed to be random. In this regard, the trained model can learn non-representative features. On the other hand, Feature Drift takes place when the pseudo-supervised and the reconstruction losses are jointly minimized. While penalizing the reconstruction loss aims to preserve all the inherent data information, optimizing the embedded-clustering objective drops the latent between-cluster variances. Due to this compromise, the clustering-friendly representations can be easily drifted. In this context, we propose ADEC (Adversarial Deep Embedded Clustering) a novel autoencoder-based clustering model, which relies on a discriminator network to reduce random features while avoiding the drifting effect. Our new metrics$\Delta _{FR}$and$\Delta _{FD}$allows to, respectively, assess the level of Feature Randomness and Feature Drift. We empirically demonstrate the suitability of our model on handling these problems using benchmark real datasets. Experimental results validate that our model outperforms state-of-the-art autoencoder-based clustering methods.