Artificial intelligence and machine learning techniques have the promise to revolutionize the field of digital pathology. However, these models demand considerable amounts of data, while the availability of unbiased t...
Artificial intelligence and machine learning techniques have the promise to revolutionize the field of digital pathology. However, these models demand considerable amounts of data, while the availability of unbiased training data is limited. Synthetic images can augment existing datasets, to improve and validate AI algorithms. Yet, controlling the exact distribution of cellular features within them is still challenging. One of the solutions is harnessing conditional generative adversarial networks that take a semantic mask as an input rather than a random noise. Unlike other domains, outlining the exact cellular structure of tissues is hard, and most of the input masks depict regions of cell types. This is also the case for non-small cell lung cancer, the most common type of lung cancer. Deciding whether a patient would receive immunotherapy depends on quantifying regions of stained cells. However, using polygon-based masks introduce inherent artifacts within the synthetic images – due to the mismatch between the polygon size and the single-cell size. In this work, we show that introducing random single-pixel noise with the appropriate spatial frequency into a polygon semantic mask can dramatically improve the quality of the synthetic images. We used our platform to generate synthetic images of immunohistochemistry-treated lung biopsies. We test the quality of the images using a three-fold validation procedure. First, we show that adding the appropriate noise frequency yields 87% of the similarity metrics improvement that is obtained by adding the actual single-cell features. Second, we show that the synthetic images pass the Turing test. Finally, we show that adding these synthetic images to the train set improves AI performance in terms of PD-L1 semantic segmentation performances. Our work suggests a simple and powerful approach for generating synthetic data on demand to unbias limited datasets to improve the algorithms’ accuracy and validate their robustness.
Tomato plants are vulnerable to several diseases, each of which may cause severe harm to the plant. These adverse conditions can significantly reduce the amount and quality of agricultural yields. In crop disease diag...
Tomato plants are vulnerable to several diseases, each of which may cause severe harm to the plant. These adverse conditions can significantly reduce the amount and quality of agricultural yields. In crop disease diagnostics, convolutional neural network (CNN) algorithms have recently materialised as the most critical area of research. Therefore, this work aims to develop a robust and lightweight CNN-based diagnostic tool for tomato leaf tissue diseases in a low-power heterogeneous embedded device. The research data consists of one category of healthy tomatoes obtained from PlantVillage and nine diseases that may affect them. We evaluated the effectiveness of the lightweight CNN diagnostic tool using conventional identification criteria, including accuracy, precision, area under the curve (AUC), f1-score, and recall, on 3,635 previously unseen test images. The outcomes are more than satisfactory, with 99.04% accuracy, 99.56% AUC, 99.00% f1-score, 98.78% precision, and 99.23% recall, respectively. The developed CNN-based diagnostic tool represents a significant advance in tomato leaf disease diagnosis, paving the way for future research to develop lightweight networks in low-power embedded devices for rapid and accurate diagnosis of plant diseases, benefiting farmers.
Graph Neural Networks (GNNs) rely on graph convolutions to exploit meaningful patterns in networked data. Based on matrix multiplications, convolutions incur in high computational costs leading to scalability limitati...
详细信息
Graph Neural Networks (GNNs) rely on graph convolutions to exploit meaningful patterns in networked data. Based on matrix multiplications, convolutions incur in high computational costs leading to scalability limitations in practice. To overcome these limitations, proposed methods rely on training GNNs in smaller number of nodes, and then transferring the GNN to larger graphs. Even though these methods are able to bound the difference between the output of the GNN with different number of nodes, they do not provide guarantees against the optimal GNN on the very large graph. In this paper, we propose to learn GNNs on very large graphs by leveraging the limit object of a sequence of growing graphs, the graphon. We propose to grow the size of the graph as we train, and we show that our proposed methodology – learning by transference – converges to a neighborhood of a first order stationary point on the graphon data. A numerical experiment validates our proposed approach.
This paper presents a low-power and low-complexity direct digital-to-RF transmitter architecture, suitable for biosensing applications. The RF front end of the transmitter is based on a ring oscillator, whose output p...
This paper presents a low-power and low-complexity direct digital-to-RF transmitter architecture, suitable for biosensing applications. The RF front end of the transmitter is based on a ring oscillator, whose output phase is modulated through the charge-to-phase mechanism using a charge injection block. Hence, the phase shift keying (PSK) modulation can be performed directly in the RF domain. Post-layout simulation results show that the transmitter is able to collect, process, and transmit sensed data with the maximum data rate of 20 Mbps and an error vector magnitude (EVM) of smaller than 3.5%, while dissipating the DC power smaller than 0.5 mW. The results demonstrate that the proposed transmitter architecture is effective for wireless biosensing applications.
Due to their stochastic nature, the increase of Renewable Energy Resources as primary sources of energy for power grids creates challenges regarding the reliability and resilience of the system. In order to combat the...
Due to their stochastic nature, the increase of Renewable Energy Resources as primary sources of energy for power grids creates challenges regarding the reliability and resilience of the system. In order to combat these obstacles, expansion of Distributed Energy Resources (DERs) and their participation in grid services is necessary. Widespread participation requires prioritizing customer privacy and addressing concerns that may arise regarding communication between DERs and Grid Service Providers. Obtaining detailed information about customers’ power consumption can lead to privacy risks that may prevent users from willingly participating in services. Anonymization of individual data is one method of privacy protection that should be explored. This paper discusses the use of the IEEE 2030.5 [1] flow reservation resources to split the operating cycles of DER load profiles into unique phases. The splitting of phases increases anonymization of DERs by making it more difficult to determine the individual characteristics of each device. We discuss the results of applying this form of anonymization to a set of simulated DER load profiles and examine the effectiveness of the anonymization through the use of a linear Support Vector Machine classifier.
In part II, an op-amp-based drive is proposed and designed. Subsequently, a very accurate model for the drive circuit and the current loop is developed as a simulation platform, while its simplified version is derived...
详细信息
Due to the non-ideality of analog components, transceivers experience high levels of hardware imperfections, like in-phase and quadrature imbalance (IQI), which manifests itself as the mismatches of amplitude and phas...
Due to the non-ideality of analog components, transceivers experience high levels of hardware imperfections, like in-phase and quadrature imbalance (IQI), which manifests itself as the mismatches of amplitude and phase between the I and Q branches. Unless proper mitigated, IQI has an important and negative impact on the reliability and efficiency of high-frequency and high-data-rate systems, such as terahertz wireless networks. Recognizing this, the current paper presents an intelligent transmitter (TX) and an intelligent receiver (RX) architecture that by employing machine learning (ML) methodologies is capable to fully-mitigate the impact of IQI without performing IQI coefficients estimation. They key idea lies on co-training the TX mapper’s and RX demapper in order to respectively design a constellation and detection scheme that takes accounts for IQI. Two training approaches are implemented, namely: i) conventional that requires a considerable amount of data for training, and ii) a reinforcement learning based one, which demands a shorter dataset in comparison to the former. The feasibility and efficiency of the proposed architecture and training approaches are validated through respective Monte Carlo simulations.
In this article, a compressive sensing-based reconstruction algorithm is applied to data acquired from a nodding multibeam Lidar system following a Lissajous-like trajectory. Multibeam Lidar systems provide 3D depth i...
详细信息
We present the application of novel additively manufactured hybrid rigid-flex Vivaldi antennas as transmit and receive elements for an ultra-wideband (UWB) frequency modulated continuous wave (FM-CW) radar. This paper...
We present the application of novel additively manufactured hybrid rigid-flex Vivaldi antennas as transmit and receive elements for an ultra-wideband (UWB) frequency modulated continuous wave (FM-CW) radar. This paper focuses on a series of measurements performed at Ku-band (12–18 GHz) to assess the effectiveness of these 3- D printed antennas within the radar loop, under thermal and vibration stresses. First, we verified the antennas' satisfactory operation down to -32 °C and inferred that thermal cycling would have minimal effect on the impulse response of the radar system. Next, we investigated the effects and bounds of vibration stress that could be sustained by these antennas and offer recommendations to support even higher vibration levels in future design iterations. Additionally, we present a succinct overview of the antenna design and the experimental setup used to perform these measurements.
Litz wire is known for its ability to minimize winding losses in high-frequency applications. However, its implementation in PCB winding poses significant challenges. This paper presents a novel PCB Litz wire concept ...
Litz wire is known for its ability to minimize winding losses in high-frequency applications. However, its implementation in PCB winding poses significant challenges. This paper presents a novel PCB Litz wire concept aimed at minimizing winding loss in high-frequency applications, specifically in the design of a solid-state transformer using PCB-winding-based technology. The proposed PCB Litz wire is designed in a circular winding configuration with curved strands, optimizing its performance. The construction method and the turn-to-turn connection are demonstrated. Factors affecting the resistance of the PCB Litz wire, including the number of strands, their width, and section, are discussed. The results show that compared to traditional PCB winding, the PCB Litz wire achieves a 30% reduction in winding loss and significantly improves current distribution, leading to enhanced thermal performance. To validate the concept's effectiveness, a prototype of a 1.6/1.05kV, 190kHz, 30-kW CLLC converter is built, demonstrating an impressive 98.8% efficiency.
暂无评论