This paper investigates a pinching-antenna (PA)-enabled cognitive radio network, where both the primary transmitter (PT) and secondary transmitter (ST) are equipped with a single waveguide and multiple PAs to facilitate simultaneous spectrum sharing. Under a general Ricean fading channel model, a closed-form analytical expression for the average spectral efficiency (SE) achieved by PAs is first derived. Based on this, a sum- SE maximization problem is formulated to jointly optimize the primary and secondary pinching beamforming, subject to system constraints on the transmission power budgets, minimum antenna separation requirements, and feasible PA deployment regions. To address this non-convex problem, a two-stage optimization algorithm is developed, in which stage 1 designs the PT/ST pinching beamforming and stage 2 updates the ST transmit power. For the PT and ST pinching beamforming optimization, the coarse positions of PA are first determined at the waveguide-level. Then, wavelength-level refinements achieve constructive signal combination at the intended user and destructive superposition at the unintended user. For the ST power control, a closed-form solution is derived. Simulation results demonstrate that i) PAs can achieve significant SE improvements over conventional fixed-position antennas; ii) the proposed pinching beamforming design achieves effective interference suppression and superior performance for both even and odd numbers of PAs; and iii) the developed two-stage optimization algorithm enables nearly orthogonal transmission between the primary and secondary networks.

Image denoising aims to enhance image quality and visual effects, finding extensive applications in various fields such as digital photography, medical imaging, video processing, and image restoration. Although deep learning-based denoising methods have made significant progress, the post-denoising results often suffer from information loss and blurring, and the network structures tend to be complex. To address these challenges, this study proposes a Multi-Scale Connected Network for Image Denoising (MSCNet) based on U-net. Specifically, this paper introduces a skip connection designed to effectively fuse shallow fine-grained information and deep semantic information, thereby improving the effects of information transmission and integration. Additionally, a channel-wise attention mechanism and a non-linear module involving point-wise multiplication of half-channel activations are presented, aiming to extract more abundant semantic features while reducing network computation. Experimental results on multiple datasets validate that, compared to other denoising methods, MSCNet demonstrates superior denoising performance.

The capacity to analyze the causes of poor decisions made by visual recognition models is becoming increasingly crucial as the security requirements of various real-world systems continue to escalate. However, the complex structure and blackbox nature within deep neural networks constrain the mining of their error causes. Based on this, we propose a concept-based (e.g. a group of pixel blocks that contain leaves represents the concept of leaves) automated strong localization interpretation framework, called hierarchically optimized concept-sensitive interpretation (HOCS), to provide quantitative analysis of the semantics of wrong decisions in the classification network is provided from two directions of internal and external information interference of samples. HOCS was applied to models with spurious correlation and well-distributed data in the training set. The results showed that it provided concrete explanations in a way that was understandable to humans and demonstrated the significant advantages of HOCS in terms of efficiency and accuracy.

Image classification models can exhibit high accuracy. However, the reasons for predictions often unable to provide explanations. Therefore, understanding the rationale behind network decisions is paramount. Numerous interpretability methods primarily focus on calculating the importance of features for output, while these features may not be precise enough to pinpoint specific regions within an image. This paper proposes a concept-similarity-based explanation method, which extracts both the main object and background concepts from images firstly, then computes the similarity between all concepts under the incorrectly predicted label and those under the correct label for samples which prediction are incorrect. This enables us to analyze if the main object or the background concepts influence the network erroneous decisions. Moreover, considering that multiple concepts could potentially lead to errors in neural networks. This article employs reinforcement learning to further reduce the set of concepts. Through the interpretation of GoogleNet trained on the ImageNet, our method demonstrates the capability to precisely identify error causes down to specific pixel block concepts.

In contrast to the intelligent reflecting surface (IRS) as a pure reflection device in the past, in this letter, from a more micro perspective, we propose to employ pseudo-random constellation rotation to secure microwave quick response (QR) code communication. Specifically, confidential information is encoded into a QR code, which is displayed on an IRS. Pseudo-random constellation rotation is applied to each element of the IRS to encrypt the QR code. This ensures that only the authenticated user (Bob) can decode and access the confidential information, while an eavesdropper (Eve) is unable to correctly interpret the QR code. We derive a closed-form expression for the bit error rate (BER) and analyze the impact of various factors on the BER for both Bob and Eve, including the distance between Bob and the IRS, Bob’s transmit power, and the Rician factor. Simulation results demonstrate the proposed scheme’s effectiveness in achieving secure communication.

This paper proposes to integrate reconfigurable intelligent surface with backscatter communication (RIS-BackCom) for downlink non-orthogonal multiple access (NOMA) networks, where a RIS serves as a backscatter device to transmit the modulated signals to multiple single-antenna target users. Building upon the established system architecture, the weighted sum rate (WSR) is maximized for all the users under the constraints of total transmit power, RIS phase shift, rate fairness, and successive interference cancellation decoding rate. By employing the techniques of Lagrangian dual transform, quadratic transform and alternative optimization strategies, the original optimization problem is decomposed into three tractable sub-problems. Then, these sub-problems are effectively addressed using successive convex approximation and semidefinite relaxation methodologies. Experimental results demonstrate the feasibility and superiority of the proposed RIS-BackCom aided NOMA system.