Pruning refers to the elimination of trivial weights from neural networks. The sub-networks within an overparameterized model produced after pruning are often called Lottery tickets. This research aims to generate win...
详细信息
A Negative Group Delay (NGD) prototype filter design based on the reciprocal transfer function of a low-pass Butterworth filter of a given order, is presented. The out-of-band gain of the prototype transfer function i...
详细信息
In order to cope with the current climate crisis, reducing greenhouse gas emissions is an important issue, and it is necessary to increase the amount of renewable energy generation. However, there is a problem of econ...
详细信息
Deep learning models for computer vision applications specifically and for machine learning generally are now the state of the art. The growth of size and complexity of neural networks has made them more and more reli...
详细信息
Deep learning models for computer vision applications specifically and for machine learning generally are now the state of the art. The growth of size and complexity of neural networks has made them more and more reliable, yet in greater need of computational power and memory as is evident from the heavy reliance on graphical processing units and cloud computing for training them. As the complexity of deep neural networks increases, the need for fast processing neural networks in real-time embedded applications at the edge also increases and accelerating them using reconfigurable hardware suggests a solution. In this work, a convolutional neural network based on the inception net architecture is first optimized in software and then accelerated by taking advantage of field programmable gate array (FPGA) parallelism. Genetic algorithm augmented training is proposed and used on the neural network to produce an optimum model from the first training run without re-training iterations. Quantization of the network parameters is performed according to the weights of the network. The resulting neural network is then transformed into hardware by writing the register transfer level (RTL) code for FPGAs with exploitation of layer parallelism and a simple trial-and-error allocation of resources with the help of the roofline model. The approach is simple and easy to use as compared to many complex existing methods in literature and relies on trial and error to customize the FPGA design to the model needed to work on any computer vision or multimedia application deep learning model. Simulation and synthesis are performed. The results prove that the genetic algorithm reduces the number of back-propagation epochs in software and brings the network closer to the global optimum in terms of performance. Quantization to 16 bits also shows a reduction in network size by almost half with no performance drop. The synthesis of our design also shows that the Inception-based classifier is cap
Beam-displacement measurements are widely used in optical sensing and communications; however, their performance is affected by numerous intrinsic and extrinsic factors, including beam profile, propagation loss, and r...
详细信息
Beam-displacement measurements are widely used in optical sensing and communications; however, their performance is affected by numerous intrinsic and extrinsic factors, including beam profile, propagation loss, and receiver architecture. Here we present a framework for designing a classically optimal beam-displacement transceiver, using quantum estimation theory. We consider the canonical task of estimating the position of a diffraction-limited laser beam after passing through an apertured volume characterized by Fresnel-number product DF. As a rule of thumb, higher-order Gaussian modes provide more information about beam displacement, but are more sensitive to loss. Applying quantum Fisher information, we design mode combinations that optimally leverage this trade-off, and show that a greater than tenfold improvement in precision is possible, relative to the fundamental mode, for a practically relevant DF=100. We also show that this improvement is realizable with a variety of practical receiver architectures. Our findings extend previous works on lossless transceivers, may have immediate impact on applications, such as atomic force microscopy and near-field optical communication, and pave the way towards globally optimal transceivers using nonclassical laser fields.
In this article, a novel method is proposed to facilitate the design of compact, low-profile, pattern reconfigurable antennas with fixed or switchable circular polarization (CP) for Internet of Vehicles (IoV) applicat...
详细信息
Co-saliency detection within a single image is a common vision problem that has not yet been well addressed. Existing methods often used a bottom-up strategy to infer co-saliency in an image in which salient regions a...
详细信息
Co-saliency detection within a single image is a common vision problem that has not yet been well addressed. Existing methods often used a bottom-up strategy to infer co-saliency in an image in which salient regions are firstly detected using visual primitives such as color and shape and then grouped and merged into a co-saliency map. However, co-saliency is intrinsically perceived complexly with bottom-up and top-down strategies combined in human vision. To address this problem, this study proposes a novel end-toend trainable network comprising a backbone net and two branch nets. The backbone net uses ground-truth masks as top-down guidance for saliency prediction, whereas the two branch nets construct triplet proposals for regional feature mapping and clustering, which drives the network to be bottom-up sensitive to co-salient regions. We construct a new dataset of 2019 natural images with co-saliency in each image to evaluate the proposed method. Experimental results show that the proposed method achieves state-of-the-art accuracy with a running speed of 28 fps.
Rate-splitting multiple access(RSMA) has recently gained attention as a potential robust multiple access(MA)scheme for upcoming wireless networks. Given its ability to efficiently utilize wireless resources and design...
详细信息
Rate-splitting multiple access(RSMA) has recently gained attention as a potential robust multiple access(MA)scheme for upcoming wireless networks. Given its ability to efficiently utilize wireless resources and design interference management strategies, it can be applied to unmanned aerial vehicle(UAV) networks to provide convenient services for large-scale access ground users. However, due to the line-of-sight(LoS) broadcast nature of UAV transmission, information is susceptible to eavesdropping in RSMA-based UAV networks. Moreover, the superposition of signals at the receiver in such networks becomes complicated. To cope with the challenge, we propose a two-user multi-input single-output(MISO) RSMA-based UAV secure transmission framework in downlink communication networks. In a passive eavesdropping scenario, our goal is to maximize the sum secrecy rate by optimizing the transmit beamforming and deployment location of the UAV-base station(UAV-BS),while considering quality-of-service(QoS) constraints, maximum transmit power, and flight space limitations. To address the non-convexity of the proposed problem, the optimization problem is first decoupled into two subproblems. Then, the successive convex approximation(SCA) method is employed to solve each subproblem using different propositions. In addition, an alternating optimization(AO)-based location RSMA(L-RSMA) beamforming algorithm is developed to implement joint optimization to obtain the suboptimal solution. Numerical results demonstrate that(1) the proposed L-RSMA scheme yields a28.97% higher sum secrecy rate than the baseline L-space division multiple access(SDMA) scheme;(2) the proposed L-RSMA scheme improves the security performance by 42.61% compared to the L-non-orthogonal multiple access(NOMA) scheme.
Photovoltaic(PV)modules age with time for various reasons such as corroded joints and terminals and glass coating defects,and their ageing degrades the PV array *** the help of the PV array numerical model,this paper ...
详细信息
Photovoltaic(PV)modules age with time for various reasons such as corroded joints and terminals and glass coating defects,and their ageing degrades the PV array *** the help of the PV array numerical model,this paper explores the effects of PV module ageing on the PV array power,and the power gains and costs of rearranging and recabling aged PV modules in a PV *** numerical PV array model is first revised to account for module ageing,rearrangement and recabling,with the relevant equations presented *** updated numerical model is then used to obtain the array powers for seven different PV *** power results are then analysed in view of the attributes of the seven PV array examples.A guiding method to recommend recabling after rearranging aged modules is then proposed,leading to further significant power gains,while eliminating intra-row *** certain conditions are met,it was shown that recabling PV modules after rearranging them may lead to further significant power gains,reaching 57%and 98%in two considered PV array *** gains are possible in other arrays.A cost-benefit analysis weighing annual power gains versus estimated recabling costs is also given for the seven considered PV array examples to guide recabling decisions based on technical and economic *** the considered examples,recabling costs can be recovered in<4 *** with the powers of the aged arrays,power gains due to our proposed rearranging and recabling the PV arrays ranged between 73%and 131%in the considered examples—well over the gains reported in the ***,the cost of our static module rearrangement and recabling method outshines the costs of dynamic reconfiguration methods recently published in the literature.
暂无评论