Array最新文献

Recently, a novel type of Neural Network (NNs): the Physics-Informed Neural Networks (PINNs), was discovered to have many applications in computational physics. By integrating knowledge of physical laws and processes in Partial Differential Equations (PDEs), fast convergence and effective solutions are obtained. Since training modern Machine Learning (ML) systems is a computationally intensive endeavour, using Quantum Computing (QC) in the ML pipeline attracts increasing interest. Indeed, since several Quantum Machine Learning (QML) algorithms have already been implemented on present-day noisy intermediate-scale quantum devices, experts expect that ML on reliable, large-scale quantum computers will soon become a reality. However, after potential benefits from quantum speedup, QML may also entail reliability, trustworthiness, safety, and security risks. To solve the challenges of QML, we combine classical information processing, quantum manipulation, and processing with PINNs to accomplish a hybrid QML model named quantum based PINNs.

A large number of sensor and actuator devices are being deployed for sensing and automation in a smart environment. While enabling communication for a large number of stations with RAW in IEEE 802.11ah, the state-of-the-art solutions for channel access are deficient in dealing with both periodic uplink and event-driven downlink actuation at the same time, as per the application’s criteria. In this paper, we propose IoT-MAC, a downlink traffic-aware Medium Access Control (MAC) protocol for automation in smart spaces. The proposed scheme uses new RAW frames to schedule downlink actuation traffic, considering the periodicity and freshness of uplink traffic. IoT-MAC identifies the periodicity of uplink traffic and schedules a frame without further contention. It then prioritizes critical downlink traffic without losing fresh uplink data. The performance analysis of the proposed scheme shows significant improvement in terms of throughput, delay, power consumption and packet loss for running different IoT applications.

In the present work the capability of convolutional neural networks to extract samples of leaves in images of tree’s canopy and detect the presence of different diseases and pests that manifest in deformation, discoloration or direct presence in the leaves, is studied. The sample obtained along with its location and sampling date, allows a mapping of the diseases in the field. This mapping capability will allow better decisions to be made when fighting these canopy diseases. An example of those are fungus and Aceria oleae in olive leaves. The study begins with the analysis of a data set generated in the laboratory and divided into healthy and faulty parts. The images were captured with a RGB and a multi-spectral with the blue, green, red, near infrared and red border spectra. They were taken in an image laboratory with a white background and led lighting. The objective was to carry out tests to determine the impact of each spectral channel and the possibility of using different types of cameras for the detection of diseases, as well as important factors to consider for its application in the field. Then, Mask rcnn R 50 FPN 3 was used to obtain segmented leaves and Fast-r cnn inception v2 to detect leaves. Then the detected or segmented leaves were classified with the Inception V3 network to determine which were healthy and which were diseased. With, the combination of these tools, it is possible to determine the disease level of an olive tree in the field.

Communication and sensing integrated networks (CSINs) refer to the ability of physical digital space perception and ubiquitous intelligent communication at the same time. These networks realize the perception and cooperative communication of multidimensional resources through the cooperative work of communication and sensing resources and have the ability of intelligent interaction and processing of new information flow. First, this study proposes the technical architecture of an enhanced CSIN (E-HFWN), studies its key technologies and performance indicators, and explains the air interface technology, including frame structure design, carrier aggregation, channel detection, physical skyline mapping, beamforming and management, resource allocation and scheduling. In the resource allocation scheme, an actor-critic reinforcement learning (RL) framework is used to divide the wireless resources. The goal is to maximize the amount of mutual information (MI) and minimize the end-to-end delay of the sensing terminal. Then, the performance of the E-HFWN is tested, including numerical simulation of wireless resource management, system peak rate, capacity, end-to-end delay and communication perception waveform sidelobe ratio. Finally, from the results of the E-HFWN index test, the E-HFWN is further enhanced on the basis of 5G mmWave. The enhanced sensing function can provide a priori information for the optimal and rapid scheduling of distributed computing power and provide richer data sources for artificial intelligence (AI) services and applications to enhance the robustness of the training model. The E-HFWN can contribute to the development of technologies related to 6G synaesthesia computing integrated networks, promote the consensus between academia and industry.

This paper presents a novel neural model of insect’s visual perception paradigm to address a challenging problem on detection of looming motion, particularly in extremely low-contrast, and highly variable natural scenes. Current looming detection models are greatly affected by visual contrast between moving target and cluttered background lacking robust and low-cost solutions. Considering the anatomical and physiological homology between preliminary visual systems of different insect species, this gap can be significantly reduced by coordinating motion and contrast neural processing mechanisms. The proposed model draws lessons from research progress in insect neuroscience, articulates a neural network hierarchy based upon ON/OFF channels encoding motion and contrast signals in four parallel pathways. Specifically, the two ON/OFF motion pathways react to successively expanding ON–ON and OFF–OFF edges through spatial–temporal interactions between polarity excitations and inhibitions. To formulate contrast neural computation, the instantaneous feedback normalization of preliminary motion received at starting cells of ON/OFF channels works effectively to suppress time-varying signals delivered into the ON/OFF motion pathways. Besides, another two ON/OFF contrast pathways are dedicated to neutralize high-contrast polarity optic flows when converging with motion signals. To corroborate the proposed method, we carried out systematic experiments with thousands of looming-square motions at varied grey scales, embedded in different natural moving backgrounds. The model response achieves remarkably lower variance and peaks more smoothly to looming motions in different natural scenarios, a significant enhancement upon previous works. Such robustness can be maintained against extremely low-contrast looming motion against cluttered backgrounds. The results demonstrate a parsimonious solution to stabilize looming detection against high input variability, analogous to insect’s capability.

Nonlinear anisotropic diffusion has attracted a great deal of attention for its ability to simultaneously remove noise and preserve semantic image features. This ability favors several image processing and computer vision applications, including noise removal in medical and scientific images that contain critical features (textures, edges, and contours). Despite their promising performance, methods based on nonlinear anisotropic diffusion suffer from practical limitations that have been lightly discussed in the literature. Our work surfaces these limitations as an attempt to create future research opportunities. In addition, we have proposed a diffusion-driven method that generates superior results compared with classical methods, including the popular Perona–Malik formulation. The proposed method embeds a kernel that properly guides the diffusion process across image regions. Experimental results show that our kernel encourages effective noise removal and ensures preservation of significant image features. We have provided potential research problems to further expand the current results.