In this paper we introduce a novel computational procedure to quantitatively investigate how image segmentation affects radiomics feature computation. Specifically, this study introduces four correlation coefficients that quantitatively assess the features' reliability in terms of quality, consistency, robustness, and instability of the features themselves. We validate our analysis in the case of an MRI-based study involving meningioma patients. The proposed approach has been intrinsically conceived for automated radiomics analysis and it is of potential interest for other imaging-driven applications in oncology.

Machine learning and artificial intelligence are two key emerging technologies in computer science that require vast amounts of data for a meaningful application. The requirement of such a large dataset is usually met by pooling data from various sources, which is often difficult to implement in practice due to strict data privacy constraints. Peer-to-Peer federated learning is a distributed machine learning paradigm with a primary goal of learning a well-performing global model by collaboratively learning a shared model at different data hubs without the need of sharing data. Due to its immense practical applications, there is a growing attention towards various challenges of efficient fed- erated learning including communication efficiency, assumptions on connectivity, data heterogeneity, enhanced privacy, etc. In this paper, we address the communication efficiency of Peer-to-Peer federated learning, modeling it using a graph theoretical frame- work. We show that one can draw from a range of graph-based algorithms to construct an efficient communication algorithm on a connected network, thereby matching the inference efficiency of centralized federated learning as well as that of a consolidated dataset. We conduct experiments with varied graph formations and sizes to validate our claims.

In remote sensing, plenty of multi-spectral images are publicly available from various landcover satellite missions. Contrastive self-supervised learning is commonly applied to unlabeled data but relies on domain-specific transformations used for learning. When focusing on vegetation, standard transformations from image processing cannot be applied to the NIR channel, which carries valuable information about the vegetation state. Therefore, we use contrastive learning, relying on different views of unlabelled, multi-spectral images to obtain a pre-trained model to improve the accuracy scores on small-sized remote sensing datasets. This study presents the generation of additional views tailored to remote sensing images using atmospheric correction as an alternative transformation to color jittering. The purpose of the atmospheric transformation is to provide a physically consistent transformation. The proposed transformation can be easily integrated with multiple channels to exploit spectral signatures of objects. Our approach can be applied to other remote sensing tasks. Using this transformation leads to improved classification accuracy of up to 6%.

Jargon words are commonly used in the communication of online communities. These words are characterized by special and implicit meanings that can only be comprehended by a small group of users, which brings challenges to community regulation and user engagement. For this problem, we present JargonFM, a framework with multiple interpretation modes for jargon understanding in online communities. JargonFM is designed based on the scientific explanation framework and supports three interpretation modes: jargon category prediction based on a Jargon Classifier, similar word identification based on a Jargon Synonym Selector, and representative text selection based on an Example Sentence Selector. We also implemented a jargon interpreter to demonstrate the usage and usefulness of our interpretation framework. Automatic and human evaluations suggest that JargonFM can explain jargon words more accurately and more efficiently than the existing interpretation methods, leading to its wide acceptance among the evaluation participants.

To keep up with the new technology modernization and the profit in shake of investors and stakeholders and importantly for the nation, and to ensure health and safety mining industry needs to approve new-age autonomous technologies and intelligent system in their field. Integration of Artificial Intelligence, Machine Learning, Internet of Things (IoT) and Automation are the keys to the 4th revolution in mining industry. This paper presents the overview of recent research upon artificial intelligence enhanced techniques and possibilities in mining operations and mining related domains. There is also a brief about the recent autonomous techniques and equipment in mining industry. Implementations and possibilities of artificial intelligence in safety and accident analysis of mining operations are sincerely detailed. Computer vision and spatial image analysis is also discussed as the recent advancement of deep learning and pattern recognition. Other mining related implementations of intelligent systems includes fragment analysis of ores, intelligent ventilation, on-site mineral processing simplification, digital twinning, mineral exploration, mineral price forecasting, mining equipment selection, post-mining land reclamation and scheduling. This paper also notes the detailed obstacles for implementing intelligent systems in mining industry.

Purpose: Robot-assisted ultrasound (US) systems could be used to provide a non-radiative three-dimensional (3D) reconstruction that can form the basis for guiding spine surgical procedures. Despite promising studies on this technology, there are few studies that offer insight into the robustness and generality of the approach by e.g. verifying performance in various testing scenarios. Therefore, this study provides an assessment of the robot-assisted US system, with technical details from experiments starting at the bench-top up to the pre-clinical study. Methods: A semi-automatic control strategy is proposed to ensure continuous and smooth scanning. Next, a method based on U-Net was developed to automatically process the anatomic features and derive a high quality 3D US reconstruction. Experiments on phantoms and human cadavers validate the proposed approach. Results: Average deviations of scanning force were found of 2.84 ± 0.45 N on synthetic phantoms and of 5.64 ± 1.10 N on human cadavers. The reconstruction yielded a mean 3D representation error of 1.28 ± 0.87 mm for the synthetic phantoms and of 1.74 ± 0.89 mm for the human cadavers. Conclusion: These results show the proposed system has high safety performance and could keep the probe force stable within the range of 1 N to obtain a good reconstruction. The experiments indicate that the proposed robotassisted US approach works in quite different experimental settings. The results encourage to also explore the potential of this technology in in-vivo studies.

A source-degenerated Schmitt-trigger logic is proposed. Based on this architecture, basic logic-gate architectures, ring oscillators, and edge-sensitive D-type flip-flops were fabricated using low-voltage p-channel organic thin-film transistors and integrated thin-film carbon resistors on flexible polymeric substrates. Mathematical analysis of this logic architecture indicates improved stability against charge-carrier-mobility degradation. A small-signal gain of 83 dB and a noise margin of 92 % of half the supply voltage have been measured. Owing to the intrinsic positive-feedback loop of the source-degenerated Schmitt-trigger logic, the design of the D-flip-flop enables a substantially smaller footprint and a considerably lower energy dissipation compared to previously reported unipolar organic-TFT-based D-flip-flop architectures.

Recurrent Neural Networks (RNN) are widespread for Video Super-Resolution (VSR) because of their proven ability to learn spatiotemporal inter-dependencies across the temporal dimension. Despite RNN’s ability to propagate memory across longer sequences of frames, vanishing gradient and error accumulation remain major obstacles to unidirectional RNNs in VSR. Several bi-directional recurrent models are suggested in the literature to alleviate this issue; however, these models are only applicable to offline use cases due to heavy demands for computational resources and the number of frames required per input. This paper proposes a novel unidirectional recurrent model for VSR, namely “Replenished Recurrency with Dual-Duct” (R2D2), that can be used in an online application setting. R2D2 incorporates a recurrent architecture with a sliding-window-based local alignment resulting in a recurrent hybrid architecture. It also uses a dual-duct residual network for concurrent and mutual refinement of local features along with global memory for full utilisation of the information available at each timestamp. With novel modelling and sophisticated optimisation, R2D2 demonstrates competitive performance and efficiency despite the lack of information available at each time-stamp compared to its offline (bi-directional) counterparts. Ablation analysis confirms the additive benefits of the proposed sub-components of R2D2 over baseline RNN models.The PyTorch-based code for the R2D2 model will be released at R2D2 GitRepo.

Society depends on a secure electricity supply which is increasingly important as various sectors are being electrified to substitute fossil fuels. The sustainability targets for the future cannot be met unless the necessary transformation of the power system is addressed. The future power system should be renewable, flexible, intelligent, secure, and resilient. This transformation should be made at acceptable costs, without jeopardising the security of electricity supply. The electricity grid is a crucial enabler for the energy transition. Intensified electrification and digitalisation imposes new challenges and requirements for the grid.  At the same time, the power system will be more exposed to natural disasters in the future due to climate change and increased weather-related stress. It is important to understand the new risk picture and find the right strategies for dealing with the security of electricity supply in the future power system.

In this paper, the design and implementation of an multibeam metasurface antenna with multiple independent channels using orbital angular momentum (OAM) demultiplexing feeding are proposed. Several aperture phase distributions that convert incident vortex beams with different and orthogonal topological orders into designated directional pencil beams are acquired and superimposed to form a metasurface aperture that, under excitations of vortex beams with multiple orders, produces independent multibeam each with a predefined direction. For implementation, a transmissive metasurface working at 5.15 GHz is designed with four beams pointing at arbitrary half space directions corresponding to incident vortex beams of order +1, −1, +3 and −3 respectively. Techniques such as OAM modes spacing and aperture partitioning are also utilized for reducing sidelobe levels (SLLs). Furthermore, a similar multibeam metasurface antenna with additional linear-to-circular polarization conversion function is proposed, which opens the possibility of utilizing different polarizations on top of the OAM based multibeam. The proposed multibeam metasurface antenna based on OAM demultiplexing feeding provides a low-cost solution for independent channels multibeam realization compared with phased arrays and realizes flexible multibeam directions compared with multibeam antennas based on lenses or reflectors.

In conjunction with the climate summit held in Sharm El Sheikh, Arab Republic of Egypt. We present to you this research paper specialized in creating an artificial neural network to predict the global average monthly temperature. We chose this topic for our research paper because of its great importance to humanity in general with its different beliefs, religions, ethnicities and political orientations. The increase in average temperatures on Earth is one of the most visible effects of global warming, and the National Oceanic and Atmospheric Administration (NOAA) indicates that the average global temperature has risen by about 0.8 degrees Celsius over the past 100 years only, and the National Oceanic and Atmospheric Administration and NASA announced 2016 was the hottest year since record keeping began in 1895, with an average temperature increase of 0.99°C above the average temperature for the entire century. We are not weather specialists and we don’t have much knowledge in this field, but we are technology and programming specialists in particular. Therefore, as a contribution to spreading awareness about the increasing global temperature, we designed and programmed an artificial neural network to predict the monthly global average temperature.

To utilize the context information, this paper proposes a novel successive POI recommendation method, SQPMF, which integrates user personal preference, user social relationship and POI transition relationship into the system for accurate recommendation of the next POI. Our experimental evaluation using two real-life datasets, Foursquare and Gowalla, show that our method SQPMF consistently outperforms all state-of-the-art systems in recommendation of successive POIs. Compared with other methods, SQPMF improves Precision and Recall by more than 21.23% and 22.71%, respectively.

Sound source localization is used in various applications such as industrial noise-control, speech detection in mobile phones, speech enhancement in hearing aids and many more. Newest video conferencing setups use sound source localization. The position of a speaker is detected from the difference in the audio waves received by a microphone array. After detection the camera focuses onto the location of the speaker. The human brain is also able to detect the location of a speaker from auditory signals. It uses, among other cues, the difference in amplitude and arrival time of the sound wave at the two ears, called interaural level and time difference. However, the substrate and computational primitives of our brain are different from classical digital computing. Due to its low power consumption of around 20 Watts and its performance in  real time the human brain has become a great source of inspiration for emerging technologies. One of these technologies is neuromorphic hardware which implements the fundamental principles of brain computing identified until today using \ac{CMOS} technologies and new devices. In this work we propose the first neuromorphic closed-loop robotic system that uses the interaural time difference for sound source localization in real time. Our system can successfully locate sound sources such as human speech. In a closed-loop experiment, the binaural robotic platform turned immediately into the direction of the sound source with a turning velocity linearly proportional to the angle difference between sound source and pan-tilt unit. After this initial turn, the robotic platform remains at the direction of the sound source. Even though the system only uses very few resources of the available hardware and was only tuned by hand it already reaches performances comparable to other neuromorphic approaches. The sound source localization system presented in this article brings us one step closer towards neuromorphic event-based systems for robotics and embodied computing.

This paper presents a new efficient and robust method for the automatic design of Metasurface (MS) by using a Deep Neural Network (DNN), which has less complexity and design time compared to conventional MS design methods. The main contribution of the present method is its ability the design reflective/transmissive MS for wave manipulation in a wide range of frequencies (12 GHz - 18 GHz). To this end, several multi-bit encoded MSs are provided that are of potential interest in single/multi-beam, Orbital Angular Momentum (OAM), and Phase Gradient Metasurface (PGM) antenna applications. In each case, the proposed DNN provides the required unit cell in two transmission and reflection modes from 12 GHz to 18 GHz for different targets. In order to demonstrate the performance of the proposed DNN method, the following steps are implemented: (i) first, the architecture of the DNN method and the creation of pixelated unit cells by DNN are discussed. This step presents 3-bit coded unit cells to design Intelligent Reflective Surfaces (IRS) in wireless communication systems. (ii) several single/multibeam reflective OAM and transmissive PGM high gain antenna structures are presented. (iii) In order to validate the proposed design concept, a Transmissive PGM lens antenna prototype with 22.6 dBi gain and Aperture Efficiency (AE) of 40% is experimentally tested at 14 GHz. Simulation and measurement results are in good agreement, which confirms the accuracy of the proposed method. The proposed DNN method can be considered a new reliable approach for designing reflective/transmissive MSs in applications of antenna and microwave with their highly demanding market.

Multispectral Imaging (MSI) collects a datacube of spatio-spectral information of a scene. Many acquisition methods for spectral imaging use scanning, preventing its widespread usage for dynamic scenes. On the other hand, the conventional color filter array (CFA) method often used to sample color images has also been extended to snapshot MSI using a Multispectral Filter Array (MSFA), which is a mosaic of selective spectral filters placed over the Focal Plane Array (FPA). However, even state-of-the-art MSFAs coding patterns produce artifacts and distortions in the reconstructed spectral images, which might be due to the non-optimal distribution of the spectral filters. To reduce the appearance of artifacts and provide tools for the optimal design of MSFAs, this paper proposes a novel mathematical framework to design MSFAs using a Sphere Packing (SP) approach. By assuming that each sampled filter can be represented by a sphere within the discrete datacube, SP organizes the position of the equal-size and disjoint spheres’s centers in a cubic container. Our method is denoted Multispectral Filter Array by Optimal Sphere Packing (MSFA-OSP), which seeks filter positions that maximize the minimum distance between the spheres’s centers. Simulation results show an image quality improvement of up to 2 dB and a remarkable boost in spectral similarity when using our proposed MSFA design approach for a variety of reconstruction algorithms. Moreover, MSFA-OSP notably reduces the appearance of false colors and zipper effect artifacts, often seen when using state-of-the-art demosaicking algorithms. Experiments using synthetic and real data prove that the proposed MSFA-OSP outperforms state-of-the-art MSFAs in terms of spatial and spectral fidelity.

The standard interpretation of Young’s double slit experiment is that the light behaves as waves before and after passing through the diaphragm of the double slit. There is one primary maximum intensity at the central position of the interference pattern. Recently, the photon chamber has been proposed and utilized to study the behavior of photons near the detecting screen. The experiments show that the light, near the screen, behaves as photons propagating along the straight-line trajectories that distribute as waves. In this article, we place the photon chamber near the diaphragm of the double slit. The Experiments show, the first time, that (1) within certain distance from the diaphragm, there is the non-wave pattern (two photon tracks), which indicates that the standard double slit experiments violate the wave interpretations; (2) the light beams behave as photons near the double slit.

Due to the necessity for highly efficient and compact power electronic converters to operate at higher frequencies, traditional wire-wound magnetics are not suitable. A comprehensive review of planar magnetic technologies is provided in this paper, which discusses the advantages as well as the disadvantages that can be associated with them. An extensive review of research literature is presented with the purpose of suggesting models for planar magnetics. Several strategies are presented to overcome the shortcomings of planar magnetics, such as winding conduction loss, leakage inductance, and winding capacitance. The purpose of this study is to provide engineers and researchers with a clear roadmap to designing planar magnetic devices.

Efficient and ultrafast (aluminum)-gallium-nitride-based metal-semiconductor-metal (Alx)Ga1–xN ultraviolet photodiodes were implemented to investigate the spectral properties of GaN/AlGaN-based photodetectors. Al composition of the GaN templates were varied from 0-% to 30% to demonstrate the impact of Al mole fraction on the cutoff wavelength of these UV photodiodes. Asymmetric metal contact electrodes were fabricated to optimize the external quantum efficiency without compromising their ultrafast photoresponse. The best performing devices exhibited a peak response of 4 V and an oscilloscope limited temporal pulse width of 61 ps at a 20-V bias with a peak spectral responsivity of 36.5 mA/W at 280 nm.