Multi-view clustering aims to group data by integrating complementary information from multiple views. However, the inherent heterogeneity among views often leads to feature entanglement, severely limiting clustering performance. To address this challenge, we propose DC-SPAN—a Dual Contrastive Attention Network—grounded in a disentangle-then-fuse paradigm. DC-SPAN employs a dual-path variational architecture to explicitly decompose each view into shared and private latent subspaces. These representations are then robustly integrated via a Product-of-Experts (PoE) mechanism. At the heart of our model is a novel dual contrastive learning objective that simultaneously encourages alignment of shared components across views and enforces separation of private ones, enabling structured and disentangled representations. A gated attention fusion module further adaptively aggregates these latent factors to yield a unified, discriminative embedding. The overall model is trained end-to-end using a composite loss function that incorporates reconstruction, orthogonality, and contrastive terms, along with a two-stage training scheme for improved stability. Extensive experiments on benchmark datasets demonstrate that DC-SPAN consistently outperforms existing state-of-the-art methods, highlighting its effectiveness and robustness in handling multi-view heterogeneity.

Modern deep learning models, such as Relation Graph Convolutional Network (RGCN), Sparse Convolutional Networks (SpConv), and Mixture of Experts Networks (MoE), are significantly dependent on the (gather-matmul-scatter) (abbreviated as (g-mm-s) ${ }_{\mathrm{s}}$) workload as their fundamental computational pattern. While existing works have made optimization attempts, several critical challenges remain unsolved, including domain-specific optimization migration, time-consuming exploration, and inefficient dataflow with dynamic inputs.To address these challenges, we introduce Efficient-GMS, a comprehensive framework that enhances ($\mathrm{g}-\mathrm{mm}-\mathrm{s})_{\text {s }}$ workload across diverse input scenarios. Our framework introduces (1) A Fusion-aware framework enabling cross-model optimization migration. We propose a comprehensive dataflow analysis that identifies shared computational patterns across models, enabling the development of four optimized dataflow patterns with vertical and horizontal fusion strategies. (2) Performance model-guided configuration space reduction. We develop a performance model to predict the relative execution efficiency across configurations, thereby reducing the search space and minimizing search time while ensuring optimal configuration selection. (3) Adaptive dataflow selection mechanism. We implement a lightweight heuristic model that dynamically selects optimal dataflow patterns based on the characteristics of the input and the hardware. Experimental results demonstrate that Efficient-GMS achieves significant performance gains, delivering an average end-to-end speedup of $1.46 \times$ in RGCN model, $1.32 \times$ in Sp-Conv-based model, and $1.15 \times$ in MoE model compared to state-of-the-art methods.

Large language models (LLMs) achieve remarkable performance across diverse tasks but face increasing GPU-memory demands due to the growing variety and complexity of downstream tasks. Efficient inference has thus become essential, especially for resource-limited settings. In this paper, we propose LoRA Decompose, a novel compression approach based on a key insight: instruction-fine-tuned models share a common pretrained-like base component and differ primarily through low-rank, LoRA-like delta components. Leveraging this observation, we reformulate the inference problem as a constrained optimization task that jointly identifies a shared low-rank structure across multiple models, significantly reducing their memory footprints. We solve this optimization efficiently using a custom-designed block coordinate descent algorithm, converging quickly within a few iterations. Empirical experiments with Llama-2 7B and 13B models demonstrate that our method achieves a remarkable >32x GPU memory reduction while preserving task accuracy, allowing substantial efficiency gains for practical deployment.

Recently, the Physics-encoded Recurrent Convolutional Neural Network (PeRCNN) has garnered significant attention for solving partial differential equations (PDEs) using deep learning methods. It acts as a discrete learning model to force encoding a given physical structure in a recurrent convolutional neural network, which outperforms other methods such as Physics-Informed Neural Network (PINN). However, PeRCNN often struggles to converge when solving PDEs with large time steps. To address this limitation, we propose an enhanced approach, Physics-encoded AutoRegressive Convolutional Neural Network with AttentionNet (A-PeARCNN), which integrates Coordinate Attention and AttentionNet within a spatiotemporal autoregressive architecture, while incorporating historical information through a sliding window. Experimental results show that A-PeARCNN improves solution accuracy by approximately 30% compared to the baseline PeRCNN, demonstrating the effectiveness of the proposed method and extending the capabilities of PeRCNN.

In this paper, we focus on the multi-depot vehicle routing problem with time window constraints (MDVRPTW) under uncertain demands. It constitutes a practical and challenging problem in logistics and supply chain management. When customers’ demands change, using the deterministic VRP model may turn the originally feasible routing solution into an infeasible one, which eventually will lead to profit loss. To address this issue, we propose a robust model to minimize the total cost while considering the uncertainty of customers’ demands. We then convert the model into a linear robust formulation and solve it using Gurobi. The experimental results verify that the robust optimization model is more reliable and resilient than the deterministic one, and can effectively suppress the perturbations caused by the demands uncertainty.

Running out of GPU memory has become a main bottleneck for large-scale DNN training. How to reduce the memory footprint during training has received intensive research attention. We find that previous gradient accumulation reduces activation memory but fails to be compatible with gradient memory reduction due to a contradiction between preserving gradients and releasing gradients. To address this issue, we propose a novel optimizer accumulation method for Adam, named Adam Accumulation (AdamA), which enables reducing both activation and gradient memory. Specifically, AdamA directly integrates gradients into optimizer states and accumulates optimizer states over micro-batches, so that gradients can be released immediately after use. We mathematically and experimentally demonstrate AdamA yields the same convergence properties as Adam. Evaluated on transformer-based models, AdamA achieves up to 23% memory reduction compared to gradient accumulation with less than 2% degradation in training throughput. Notably, AdamA can work together with memory reduction methods for optimizer states to fit 1.26×~3.14× larger models over PyTorch and DeepSpeed baseline on GPUs with different memory capacities.

This paper investigates a scheduling problem with due times for unrelated parallel machines, incorporating uncertain setup times and processing times. The tardiness of jobs can be easily overlooked in production scheduling problems when dealing with uncertainty, potentially leading to disruptions in the production plan and longer production cycles. To address this issue, we propose a robust model to minimize the total tardiness while considering the uncertainty of setup times and processing times, as well as incorporating the constraint of due times. We then convert the model into a linear robust formulation and solve it using Gurobi. Through experimental comparisons between the deterministic model and the proposed robust model under uncertain conditions, the results demonstrate that the robust model exhibits superior performance in handling uncertain situations.

Abstract Owing to the renaissance of deep convolutional neural networks (CNN), salient object detection based on fully convolutional neural networks (FCNs) has attracted widespread attention. However, the scale variation of prominent objects, complex background features and fuzzy edges have historically been a great challenge to us. All these are closely associated with the utilization of multi‐level and multi‐scale features. At the same time, deep learning methods meet the challenges of computation and memory consumption in practice. To address these problems, the authors propose a different salient object detection method based on residuals learning and dense fusion learning framework. The proposed network is named Residual Dense Collaborative Network (RDCNet). First of all, the authors design a multi‐layer residual learning (MRL) module to extract salient object features in more detail, getting the utmost out of the object's multi‐scale and multi‐level information. Then, on the basis of the vigoroso stage‐wise convolution feature, the authors put forward the dilated convolution module (DCM) to acquire a rough global saliency map. Finally, the final accurate saliency detection map is obtained through dense cooperation learning (DCL), and the remaining learning is also used to improve gradually, so as to achieve high compactness and high‐efficiency results. Experimental results show that this method is the most advanced method for five widely used datasets (DUTS‐TE, HKU‐IS, PASCAL‐S, ECSSD, DUT‐OMRON) without any pre‐processing and post‐processing. Especially on the ECSSD dataset, the F‐measure of RDCNet achieves 95.2%.

Graph Convolutional Network (GCN) is a promising but computing- and memory-intensive learning model. Processing-in-memory (PIM) architecture based on the ReRAM crossbar is a natural fit for GCN inference. It can reduce the data movements and compute the vector-matrix multiplication (VMM) in analog. However, it requires an unbearable crossbar cost to leverage the massive parallelism exhibited in GCNs. This paper explores the design space for GCN acceleration on ReRAM crossbars and presents the first PIM-based GCN accelerator named PIMGCN. PIMGCN employs dense data mapping and a search-execute architecture to take full advantage of the intra-vertex parallelisms with acceptable crossbars cost. We further propose two scheduling strategies for PIMGCN to maximize the inter-vertex parallelisms and optimize the pipeline. The optimal scheduling is reduced to a maximum independent set problem, which is solved by a novel node-grouping algorithm. Compared to the state-of-the-art software framework running on Intel Xeon CPU and NVIDIA RTX8000 GPU, PIMGCN achieves on average 11044× and 74.3× speedup, 6.13E+06× and 5.09E+03× energy reduction, respectively. Compared with ASIC accelerator HyGCN [1], PIMGCN achieves 219× speedup and 95.3× energy reduction.

Abstract The road extraction from high resolution remote sensing image is of great importance in a variety of applications. Recently, the abundant deep convolutional neural networks are proposed for road extraction task. However, the existing approaches lack suitable strategy to utilize multiple views road features for road extraction, which fails to extract road with smooth appearance and accurate boundary under complex scenes. To address this problem, the authors propose a novel deep residual and pyramid pooling network (DRPPNet) for extracting road regions from high resolution remote sensing image. The DRPPNet consists of three parts: deep residual network (DResNet), pyramid pooling module (PPM) and deep decoder (DD). Specially, the DResNet uses several residual blocks to extract deep road features from input images, which can enhance learning ability of DRPPNet and avoid gradient vanish. Then, PPM is proposed to fuse road features from multiple views and it aims to address disadvantage of single view feature. Finally, the DD is used to recover size of feature maps to input size. Extensive experiments on two challenging road datasets demonstrate that proposed method outperforms the state‐of‐the‐art methods greatly on performance of road extraction task.

At present, artificial intelligence technology is widely used in society, and various intelligent systems emerge as the times require. Due to the uniqueness of biometrics, most intelligent systems use biometric-based recognition technology, among which face recognition is the most widely used. To improve the security of intelligent system, this paper proposes a face authentication system based on edge computing and innovatively extracts the features of face image by convolution neural network, verifies the face by cosine similarity, and introduces a user privacy protection scheme based on secure nearest neighbor algorithm and secret sharing homomorphism technology. The results show that when the threshold is 0.51, the correct rate of face verification reaches 92.46%, which is far higher than the recognition strength of human eyes. In face recognition time consumption and recognition accuracy, the encryption scheme is basically consistent with the recognition time consumption in plaintext state. It can be seen that the security of the intelligent system with this scheme can be significantly improved. This research provides a certain reference value for the research on the ways to improve the security of intelligent system.

This paper presents a novel unsupervised polarimetric synthetic aperture radar (PolSAR) image classification method, which incorporates polarimetric image factorization and deep convolutional networks into a principled framework. To implement this idea, we design a convolutional neural network (CNN) with a newly defined loss function which measures the probability distribution distance between the initial distribution maps and CNN predictions. In the proposed method, we firstly execute polarimetric image factorization to generate a dictionary of meaningful atom scatters and their corresponding distribution maps, where the strongest scatters are selected as training samples for CNN. Next, we train the CNN by iteratively optimizing the defined energy function, producing the final distribution maps and classification result. The proposed approach is applied on a real UAVSAR image. Experimental results justify that our approach can effectively classify the PolSAR image in an unsupervised way and produce favorable classification results.

Aiming at improving the classification performance with greatly reduced annotation cost, this paper presents an active deep learning approach for minimally-supervised PolSAR image classification, which integrates active learning and fine-tuning convolutional neural network (CNN) into a principled framework. Starting from a CNN trained using a very limited number of labeled pixels, we iteratively and actively select the most informative candidates for annotation, and incrementally fine-tune the CNN by incorporating the newly annotated pixels. Moreover, to boost the performance and robustness of the proposed method, we employ Markov random field to enforce label smoothness, and data augmentation technique to enlarge the training set. Extensive experiments demonstrated that our approach achieved state-of-the-art classification results with significantly reduced annotation cost.