System partitioning for effective simulation of civil infrastructure flow networks on parallel processors is a nontrivial problem. Arbitrary partitioning focused only on balancing processor workload can lead to a large interprocessor communication burden that limits parallel speedup. Thus, there is a need for intelligent partitioning algorithms that balance the estimated computational load while minimizing the number of connections between partitions. Graph theory provides widely used partitioning methods, but these are applicable to networks with power-law connectivity and where the computational workload is proportional to the number of system nodes—conditions that do not hold for finite-volume solution of water drainage networks (e.g., river systems, stormwater drainage systems). This paper presents the novel BIPquick algorithm, which is shown to be an effective approach to identifying network partitions with reduced connectivity for systems that are directed acyclic graphs (DAGs) and have a physical limit on the number of connections per network node. Novel developments include (1) a node-cut approach that allows a partitioning workload function to be exactly balanced in systems where the computational work is proportional to the link length between nodes, (2) a finite-pass approach to partitioning that ensures a partitioning solution in a known time, and (3) a new connectivity scaling metric that allows simple evaluation and comparison of different partitioning results. The BIPquick model is tested on a large river network with up to 10,000 partitions.

Image texture analysis has for decades represented a promising opportunity for cancer assessment and disease progression evaluation, evolving over time in a discipline, i.e., radiomics. However, the road for a complete translation into clinical practice is still hampered by intrinsic limitations. As purely supervised classification models fails in devising univocal imaging- based differences in tumors with different prognosis, cancer sub- typing approaches would benefit from the employment of distant supervision, for instance exploiting survival/recurrence information. In this work, we transfer our previous model for Hodgkin Lymphoma subtyping to a multi-center study case. We evaluated model performance in two independent datasets coming from two hospitals, comparing and analyzing the results. Our preliminary data confirmed the instability of radiomics due to across-center lack of reproducibility, leading to meaningful results in one center and poorer performance in the other. We then learnt stratification rules from the first dataset via Random Forest and leveraged those rules to transfer the stratification policy onto the second dataset. In this way, on the one hand, we tested the stratification ability of cancer subtyping in a validation setting and, on the other hand, enriched the noisier dataset with valuable information, in a borrowing strength fashion. The transfer of the model resulted successful. Moreover, having extracted decision rules for cancer subtyping, we were able to draw up risk factors to be considered in clinics. The work shows the potentialities of the proposed pipeline to be further evaluated in larger multi-center datasets, with the goal of translating radiomics into clinical practice.

A new algorithm for simplifying Boolean logic expressions is presented. It is a top-down guided search algorithm.

The paper considers a multi-source relaying status update system with stochastic arrivals and develops three scheduling policies to minimize the sum average age of information subject to transmission capacity and long-run average resource constraints. Namely, we provide these policies under two different scenarios regarding the knowledge of system statistics: known environment and unknown environment. For the known environment, a constrained Markov decision process approach and a drift-plus-penalty method are proposed, and for the unknown environment, a deep learning policy is developed. The structure of an optimal policy is analyzed, and simulation results are provided.

The deployment of new airborne small cells in conjunction with an existing network equipped with a multiple-input multiple-output (MIMO) system can be a promising solution to meet the high throughput and coverage needs of future networks. In this study, we present the potential benefits of deploying a large number of antenna elements in downlink vertical heterogeneous cellular networks (VHCNets). The network model consists of a terrestrial base station (TBS) with an array of massive antenna elements and an aerial base station (ABS) with a single antenna to provide connectivity for ground user equipment (GUEs). The locations of the TBS and ABS are modeled as two-dimensional and three-dimensional Poisson point processes, respectively. An appropriate air-to-ground channel model encompasses both line-of-sight (LOS) and non-line-of-sight (NLOS) communication between GUEs and ABS. Using the concepts of stochastic geometry, the performance of the VHCNet was analyzed in terms of coverage probability, area spectral efficiency, and average ergodic rate. The approximated expressions for these performance metrics were validated using Monte Carlo simulations, and a close match was observed between the analytical and simulated results. Based on the simulation results, subsequent use of a massive number of antennas at TBSs in conjunction with ABSs in the VHCNet can improve the overall performance of the network.

The IEEE 802.11ax standard is designed to provide high-efficiency WLAN operating in dense deployment, with a focus on increasing robustness and uplink transmission.  IEEE 802.11ax aims to deliver self-configuration and self-adaptation capabilities for interference management techniques in dense deployments, to improve overall network conditions. The IEEE 802.11ax medium strives to intelligently use network allocation vectors (NAV), which block simultaneous transmission of neighboring co-existing access points from transmitting for a set period of time. Given an acceptable collision probability, this collision avoidance strategy tends to degrade network throughput and increase transmission delays since the medium is never completely exploited. In this paper, we propose affinityGNN, a graph neural network approach to actively optimize network co-existence transmission in a dense scenario. We introduce direct-affinityGNN which considers direct communications between IEEE802.11ax-compliant wireless systems and in contrast, skip-affinityGNN, which considers communication between IEEE802.11ax and Legacy devises by aggregating neural messages from direct and skip (two-hop) communicating neighbors in the dense wireless network deployment. Extensive experimental results not only show the superior performance of our proposed model over the state-of-the-art, but also demonstrate its potentially good interpretability and robustness for wireless network structure.

Mammography is a popular diagnostic imaging procedure for detecting breast cancer at an early stage. Various deep learning (DL) approaches to breast cancer detection incur high costs and are prone to classify incorrectly. Therefore, they are not sufficiently reliable to replace existing techniques used by medical practitioners. Specifically, these DL approaches do not exploit the complex texture patterns and interactions in mammograms. These DL approaches need labelled data to enable learning, limiting the scalability of these methods because of sufficient labelled datasets. Further, these DL models lack generalisation capability to newly-synthesised patterns/textures. To address these problems, in the first instance, we design a graph model to transform the mammogram images into highly correlated multi-graphs, encoding rich structural relations and high-level texture features. Then, we consider a self-supervised learning multi-graph encoder (SSL-MG) to improve the features presentation, especially when limited labelled data is available. Finally, we design a semi-supervised mammogram multi graph convolution neural network downstream model (MMGCN) to perform multi classification of mammogram segments encoded in the multi-graph nodes. We evaluate the classification performance of MMGCN independently and with integration with SSL-MG in a model called SSL-MMGCN over several training settings. Our results reveal the efficient learning performance of SSL-MNGCN and MMGCN with 0.97 and 0.98 AUC classification accuracy in contrast to the multitask deep graph (GCN) method of Hao Du et al. (2021) with 0.81 AUC accuracy.

Deteriorated power system components have a higher probability of failure than new components. Still, reliability of supply analyses traditionally model all components of the same type with the same probability of failure, and thus neglects the effect of deteriorated components. This paper presents a methodology to integrate a condition dependent component probability of failure model into a power system reliability analysis. The component state is described by a semi-Markov process, and the paper shows how this, under reasonable assumptions, can be approximated by a Markov process. The Markov assumption simplifies the analysis and allows the model to include preventive retirement and be calibrated to statistical data. A case study using statistical data for Norwegian power transformers shows that, in the Norwegian power system, the proportion of failures that are due to poor condition is small, partly due to the common strategy of preventive retirement. However, if the condition of the transformers were worse, the impact of poor condition can be considerable. The methodology further enables identification of the transformers that contribute most to the risk to reliability of supply. The paper thus highlights the importance of accounting for component condition in strategic decisions such as long-term renewal planning.

This paper presents a novel solution to the inverse Frobenius-Perron problem of reconstructing an unknown nonlinear and ergodic map from causal sequences of probability density functions generated by the map. The original solution to this problem successfully reconstructs members of the canonical map class (i.e., a subset of the piecewise linear semi-Markov maps), provided all the map’s branches are monotonically increasing. The original solution constructs a matrix estimate of the map’s Frobenius-Perron operator, which governs the evolution of density functions under iteration of the map, from the density sequences. The one-dimensional map is reconstructed from this matrix. In contrast, the proposed solution constructs a higher-order matrix estimate of the Frobenius-Perron operator. A member of the newly proposed class of generalized hat maps, a superset of the canonical maps, is constructed from this matrix estimate. The proposed solution successfully distinguishes between increasing and decreasing map branches and enlarges the class of maps that can be successfully reconstructed to canonical maps with any subset of decreasing branches. When used to reconstruct any piecewise linear semi-Markov map, the proposed solution generates a map with consistent invariant density and power spectrum mode characteristics, regardless of the unknown map’s canonicity or branch monotonicity. Numerical examples that illustrate the proposed solution’s characteristics are presented.

Large-scale classification poses great challenges to  storage and computation. There are two major solutions to address the problem:  data compression and  quantization. In the paper, we study the method of first reducing data dimension by random projection and then quantizing the projections to ternary and binary codes, which has been widely applied in practice. Often, the extreme quantization would  degrade the accuracy of classification due to high quantization errors. Interestingly, however, we observe that  the quantization could result in   performance improvement, rather than degradation, if the data for quantization are preprocessed by sparse transform.  Also,  the quantization gain could be obtained with the random projections of the data, if both the data and random projection matrices are  sparse enough, such that the resulting projections remain sparse.  The intriguing  performance is verified and analyzed  with extensive experiments.

Sparse array (SA) geometries are proposed to obtain low mutual coupling (MC) and improved degrees-of-freedom (DOF), which are composed of concatenations of multiple uniform linear arrays (ULAs) with certain closed-form expressions and are able to provide long virtual difference coarrays (DCAs) than many existing SAs. Meanwhile, the new proposed SAs have low coupling leakage, hence the proposed SAs have good trade-off between DOF and MC. Numerical experiments are created to give a verify the SA superiority in the presence of heavy MC.

In wireless Internet of things (IoT), the sensors usually have limited bandwidth and power resources. Therefore, in a distributed setup, each sensor should compress and quantize the sensed observations before transmitting them to a fusion center (FC) where a global decision is inferred. Most of the existing compression techniques and entropy quantizers consider only the reconstruction fidelity as a metric, which means they decouple the compression from the sensing goal. In this work, we argue that data compression mechanisms and entropy quantizers should be co-designed with the sensing goal, specifically for machine-consumed data. To this end, we propose a novel deep learning-based framework for compressing and quantizing the observations of correlated sensors. Instead of maximizing the reconstruction fidelity, our objective is to compress the sensor observations in a way that maximizes the accuracy of the inferred decision (i.e., sensing goal) at the FC. Unlike prior work, we do not impose any assumptions about the observations distribution which emphasizes the wide applicability of our framework. We also propose a novel loss function that keeps the model focused on learning complementary features at each sensor. The results show the superior performance of our framework compared to other benchmark models.

The forthcoming Surface Water and Ocean Topography (SWOT) satellite and AirSWOT airborne instrument are the first imaging radar-altimeters designed with near-nadir, 35.75 GHz Ka-band InSAR for mapping terrestrial water storage variability. Remotely sensed surface water extents are crucial for assessing such variability, but are confounded by emergent and inundated vegetation along shorelines. However, because SWOT-like measurements are novel, there remains some uncertainty in the ability to detect certain land and water classes. We study the likelihood of misclassification between 15 land cover types and develop the Ka-band Phenomenology Scattering (KaPS) scattering model to simulate changes to radar backscatter as a result of changing surface water fraction and roughness. Using a separability metric, we find that water is five times more distinct compared with dry land classes, but has the potential to be confused with littoral zone and wet soil cover types. The KaPS scattering model simulates AirSWOT backscatter for incidence angles 1-27°, identifying the conditions under which open water is likely to be confused with littoral zone and wet soil cover types. A comparison of KaPS simulated backscatter with AirSWOT observed backscatter shows good overall agreement across the 15 classes (median r2=0.76). KaPS characterization of the sensitivity of near-nadir, Ka-band SAR to small changes in both wet area fraction and surface roughness enables more nuanced classification of inundation area. These results provide additional confidence in the ability of SWOT to classify water inundation extent, and open the door for novel hydrological and ecological applications of future Ka-band SAR missions.

In this paper, Matlab-Simulink and LabView models are constructed for a new nonlinear dynamic system of equations in an eight-dimensional (8D) phase space. For fixed parameters of the 8D dynamical system, the spectrum of Lyapunov exponents and the Kaplan-York dimension are calculated. The presence of two positive Lyapunov exponents demonstrates the hyperchaotic behavior of the 8D dynamical system. The fractional Kaplan-York dimension indicates the fractal structure of strange attractors. We have shown that an adaptive controller is used to stabilize the novel 8D chaotic system with unknown system parameters. An active control method is derived to achieve global chaotic synchronization of two identical novel 8D chaotic systems with unknown system parameters. Based on the results obtained in Matlab-Simulink and LabView models, a chaotic signal generator for the 8D chaotic system is implemented in the Multisim environment. The results of chaotic behavior simulation in the Multisim environment show similar behavior when comparing simulation results in Matlab-Simulink and LabView models.

This study proposes a new method to measure the polarization charge of ferroelectric field effect transistors (FeFETs) using a pulse generator and source measurement unit (SMU) and exploiting the charge pumping (CP) principle, which is widely followed to measure the interface trap charge density (Dit) of MOSFETs. Although the pulse waveforms applied to the gate electrode were equal to that of conventional polarization charge measurement methods, the response current was measured using the SMU similar to that in CP. As the proposed method measured the average DC current on the well or source/drain terminal, it provided a reliable quantification of polarization charges with low noises even at lower applied voltages. This measurement is possible because the dipole induced charge is a fixed charge, such as an interface charge that is not moved by the interface trap. And this method conveniently measured the switching charge even for < 50ns of rising/falling time of the applied pulse. Moreover, as the read current resolution of the SMU < 5 ×10-12 A, the switching charge can be measured with a device area < 1 µm2. Thus, the proposed CP-based method proved to be a highly effective method of polarization charge measurement for three- or four-terminal devices such as FeFETs unless the measurement requires the hysteresis curve along with polarization charges.

This paper demonstrates a deep learning based methodology for the rapid simulation of planar microwave circuits based on their layouts. We train convolutional neural networks to compute the scattering parameters of general, two- port circuits consisting of a metallization layer printed on a grounded dielectric substrate, by processing the metallization pattern along with the thickness and dielectric permittivity of the substrate. This approach harnesses the efficiency of convolutional neural networks with pattern recognition tasks and extends previous efforts to employ neural networks for the simulation of parameterized circuit geometries. Training is based on full-wave simulation data generated via the Finite-Difference Time-Domain (FDTD) method over a target frequency range. To accelerate the generation of such data, we build a hybrid neural network including recursive neural network modules, to compensate numerical dispersion errors in coarse-grid FDTD. This novel dispersion compensation scheme allows us to generate accurate FDTD training data from fast, coarse-grid simulations.

A document by Varun B. Click on the document to view its contents.

Deep neural networks are commonly used for medical purposes such as image generation, segmentation, or classification. Besides this, they are often criticized as black boxes as their decision process is often not human interpretable. Encouraging the latent representation of a generative model to be disentangled offers new perspectives of control and interpretability. Understanding the data generation process could help to create artificial medical data sets without violating patient privacy, synthesizing different data modalities, or discovering data generating characteristics. These characteristics might unravel novel relationships that can be related to genetic traits or patient outcomes. In this paper, we give a comprehensive overview of popular generative models, like Generative Adversarial Networks (GANs), Variational Autoencoders (VAEs), and Flow-based Models. Furthermore, we summarize the different notions of disentanglement, review approaches to disentangle latent space representations and metrics to evaluate the degree of disentanglement. After introducing the theoretical frameworks, we give an overview of recent medical applications and discuss the impact and importance of disentanglement approaches for medical applications. Keywords: Generative Models, Disentanglement, Representation Learning, Medical Applications