Blazegraph (United States)

Recent experimental studies showed that electrically coupled neural networks like in mammalian inferior olive nucleus generate synchronized rhythmic activity by the subthreshold sinusoidal-like oscillations of the membrane voltage. Understanding the basic mechanism and its implication of such phenomena in the nervous system bears fundamental importance and requires preemptively the connectome information of a given nervous system. Inspired by these necessities of developing a theoretical and computational model to this end and, however, in the absence of connectome information for the inferior olive nucleus, here we investigated interference phenomena of the subthreshold oscillations in the reference system Caenorhabditis elegans for which the structural anatomical connectome was completely known recently. We evaluated how strongly the sinusoidal wave was transmitted between arbitrary two cells in the model network. The region of cell-pairs that are good at transmitting waves changed according to the wavenumber of the wave, for which we named a wavenumber-dependent transmission map. Also, we unraveled that (1) the transmission of all cell-pairs disappeared beyond a threshold wavenumber, (2) long distance and regular patterned transmission existed in the body-wall muscles part of the model network, and (3) major hub cell-pairs of the transmission were identified for many wavenumber conditions. A theoretical and computational model presented in this study provided fundamental insight for understanding how the multi-path constructive/destructive interference of the subthreshold oscillations propagating on electrically coupled neural networks could generate wavenumber-dependent synchronized rhythmic activity.

Solar power generation has recently been in the spotlight as global warming continues to worsen. However, two significant problems may hinder solar power generation, considering that solar panels are installed outside. The first is soiling, which accumulates on solar panels, and the second is a decrease in sunlight owing to bad weather. In this paper, we will demonstrate that the solar power generation forecasting can increase when considering soiling and sunlight information. We first introduce a dataset containing images of clean and soiled solar panels, sky images, and weather information. For accurate solar power generation forecasting, we propose a new multimodal model that aggregates various features related to weather, soiling, and sunlight. The experimental results demonstrated the high accuracy of our proposed multimodal model.

Recent work has demonstrated the Transformer model is effective for computer vision tasks. However, the global self-attention mechanism utilized in Transformer models does not adequately consider the local structure and details of images, which may result in the loss of information and local details, causing decreased estimation accuracy in gaze estimation tasks when compared to convolution or sequential stacking methods. To address this issue, we propose a parallel CNNs-Transformer aggregation network (CTA-Net) for gaze estimation, which fully leverages the advantages of the Transformer model in modeling global context while the convolutional neural networks (CNNs) model in retaining local details. Specifically, Transformer and ResNet are deployed to extract facial and eye information, respectively. Additionally, an attention cross fusion (ACFusion) Block is embedded with CNN branch, which decomposes features in space and channels to supplement lost features, suppress noise, and help extract eye features more effectively. Finally, a dual-feature aggregation (DFA) module is proposed to effectively fuse the output features of both branches with the help feature a selection mechanism and a residual structure. Experimental results on the MPIIGaze and Gaze360 datasets demonstrate that our CTA-Net achieves state-of-the-art results.

1．IntroductionBronchoscope navigation is used for assisting physicians during bronchoscopic examinations.Existing methods for depth acquisition in bronchus scene are mainly classified into two types: compute vision-based depth estimation which calculates depth information by using structure from motion (SfM) , which shows poor result 1) .Liu et al. used a dual convolutional neural network to estimate depth map from RB images by using depth information estimated from SfM 2) , however, the preparation of the training data is time consuming.To overcome the shortcomings of the aforementioned methods, we propose a method to estimate depth information by using cycle generative adversarial network (CycleGAN) 3) .This network does not need preparation of paired training data.Previous literature shows good performance to estimate realistic colonoscopic images from virtual colonoscopic images.We use this method to estimate a depth image from real bronchoscopic (RB) images.We use CycleGAN to find an RB-to-depth translator to generate the depth images from the RB images.We evaluate the generated depth images by using manually created depth

Aperture synthesis technique is able to provide high spatial resolution without requiring very large and massive real aperture. Recently, in addition to the well-known aperture synthesis radiometer, the concept of mirrored aperture synthesis has been proposed, and the experiments has been carried out to validate this technique. However, the field of view (FOV) of a mirrored aperture synthesis radiometer (MASR) has not been analyzed. In this paper, the FOV of an one-dimensional MASR is analyzed based on the symmetric convolution. A simulation and experiment are carried out to validate the derived FOV. The FOV of a two-dimensional MASR can be analyzed in the same way.

Fast, scalable, low-cost, and low-power execution of parallel graph algorithms is important for a wide variety of commercial and public sector applications. Breadth First Search (BFS) imposes an extreme burden on memory bandwidth and network communications and has been proposed as a benchmark that may be used to evaluate current and future parallel computers. Hardware trends and manufacturing limits strongly imply that many-core devices, such as NVIDIA® GPUs and the Intel® Xeon Phi®, will become central components of such future systems. GPUs are well known to deliver the highest FLOPS/watt and enjoy a very significant memory bandwidth advantage over CPU architectures. Recent work has demonstrated that GPUs can deliver high performance for parallel graph algorithms and, further, that it is possible to encapsulate that capability in a manner that hides the low level details of the GPU architecture and the CUDA language but preserves the high throughput of the GPU. We extend previous research on GPUs and on scalable graph processing on supercomputers and demonstrate that a high-performance parallel graph machine can be created using commodity GPUs and networking hardware.