Surrogate-assisted evolutionary algorithms (SAEAs) are widely used in solving computationally expensive optimization problems. The intricate nature of real-world optimization problems necessitates the development of more efficient SAEAs to effectively address these challenges. In this paper, we propose a novel algorithm, named surrogate-assisted clustering-Based evolutionary algorithm (SACBEA) for solving expensive optimization problems. SACBEA is an innovative combination of dynamic surrogate-assisted particle swarm optimization (DSAP) and clustering-based local search (CBLS) for the balance between exploration and exploitation. Specifically, the DSAP employs the top-ranked solutions in the initialization of population, using Particle Swarm Optimization (PSO) as the evolutionary operator. Concurrently, a dynamic surrogate model is utilized for fitness prediction. In addition, the CBLS employs the k-means clustering algorithm to identify the promising region, initializing the population within this area for a more focused search. And the Differential Evolution (DE) algorithm is integrated into CBLS. SACBEA switches between DSAP and CBLS based on search performance. When one search mechanism fails to find a better solution, SACBEA seamlessly transitions to the other. To verify the effectiveness of SACBEA, a test set of 13 benchmark problems is adopted to conduct comparative experiments. Experimental results demonstrate the significant advantages of SACBEA over several state-of-the-art algorithms.

AI-synthesized text and images have gained significant attention, particularly due to the widespread dissemination of multi-modal manipulations on the internet, which has resulted in numerous negative impacts on society. Existing methods for multi-modal manipulation detection and grounding primarily focus on fusing vision-language features to make predictions, while overlooking the importance of modality-specific features, leading to sub-optimal results. In this paper, we construct a simple and novel transformer-based framework for multi-modal manipulation detection and grounding tasks. Our framework simultaneously explores modality-specific features while preserving the capability for multi-modal alignment. To achieve this, we introduce visual/language pre-trained encoders and dual-branch cross-attention (DCA) to extract and fuse modality-unique features. Furthermore, we design decoupled fine-grained classifiers (DFC) to enhance modality-specific feature mining and mitigate modality competition. Moreover, we propose an implicit manipulation query (IMQ) that adaptively aggregates global contextual cues within each modality using learnable queries, thereby improving the discovery of forged details. Extensive experiments on the DGM4dataset demonstrate the superior performance of our proposed model compared to state-of-the-art approaches.

Remote sensing images often encounter various challenges arising from differences in shooting time, location, equipment, sensors, and other factors. These disparities lead to image distortion and insufficient overlap between pairs of images captured at the same location. Consequently, the accuracy of agricultural remote sensing image registration is significantly compromised. This paper proposes a robust feature matching technique called GAN-based neighborhood representation (RFM-GAN) for the intricate registration of satellite images and unmanned aerial vehicle (UAV) images. The RFM-GAN method leverages a neighborhood representation approach based on a generative adversarial network (GAN) with two discriminators. This representation enhances the distinction between true matches (inliers) and false matches (outliers). Additionally, a dissimilarity measure network employing a self-supervised training approach, eliminating the need for manual labeling, is designed to handle the multi-view transformation of satellite and UAV images. The experimental results confirm that RFM-GAN outperforms seven other state-of-the-art methods in terms of satellite image and UAV image processing.

Abstract. Parallel imaging (PI), relying on multicoils to sense [Formula: see text]-space data, is an effective technique to accelerate magnetic resonance imaging by exploiting spatial sensitivity coding of multiple coils, with an integrated compressive sensing (CS) technology to achieve higher acceleration. In this paper, we propose a novel nonconvex reconstruction model and its proximal alternating linearized minimization (PALM) algorithm for PI in a blind setting that MR image and multichannel sensitivity maps are jointly estimated, regularized by image and sensitivity regularizers. Instead of hand-crafting the image and sensitivity regularizers, we propose unrolling the PALM algorithm to be a deep network for Blind Parallel MRI, dubbed as BPMRI-Net, with two learnable subnetworks to substitute the proximal operators of the image and sensitivity regularizers. We theoretically prove the linear convergence of BPMRI-Net as an iterative algorithm, which alternately updates two variables based on the learnable proximal operators. The learned BPMRI-Net can simultaneously output the MR image and sensitivity maps from undersampled multichannel [Formula: see text]-space data even when the number of low-frequency sampling lines in the center of [Formula: see text]-space is small. Numerical results demonstrate the effectiveness of our method with state-of-the-art reconstruction accuracy.