Unsupervised Character Recognition with Graphene Memristive Synapses

Memristive devices being applied in neuromorphic computing are envisioned to significantly improve the power consumption and speed of future computing platforms. The materials used to fabricate such devices will play a significant role in their viability. Graphene is a promising material, with superb electrical properties and the ability to be produced sustainably. In this paper, we demonstrate that a fabricated graphene-pentacene memristive device can be used as synapses within Spiking Neural Networks (SNNs) to realise Spike Timing Dependent Plasticity (STDP) for unsupervised learning in an efficient manner. Specifically, we verify operation of two SNN architectures tasked for single digit (0-9) classification: (i) a simple single-layer network, where inputs are presented in 5x5 pixel resolution, and (ii) a larger network capable of classifying the Modified National Institute of Standards and Technology (MNIST) dataset, where inputs are presented in 28x28 pixel resolution. Final results demonstrate that for 100 output neurons, after one training epoch, a test set accuracy of up to 86% can be achieved, which is higher than prior art using the same number of output neurons. We attribute this performance improvement to homeostatic plasticity dynamics that we used to alter the threshold of neurons during training. Our work presents the first investigation of the use of green-fabricated graphene memristive devices to perform a complex pattern classification task. This can pave the way for future research in using graphene devices with memristive capabilities in neuromorphic computing architectures. In favour of reproducible research, we make our code and data publicly available https://anonymous.4open.science/r/c69ab2e2-b672-4ebd-b266-987ee1fd65e7.

Fast Fractal Technique using Modified Moment Features on Domain Blocks

In this research, a new technique is suggested to reduce the long time required by the encoding process by using modified moment features on domain blocks. The modified moment features were used in accelerating the matching step of the Iterated Function System (IFS). The main disadvantage facing the fractal image compression (FIC) method is the over-long encoding time needed for checking all domain blocks and choosing the least error to get the best matched domain for each block of ranges. In this paper, we develop a method that can reduce the encoding time of FIC by reducing the size of the domain pool based on the moment features of domain blocks, followed by a comparison with threshold (the selected threshold based on experience is 0.0001). The experiment was conducted on three images with size of 512x512 pixel, resolution of 8 bits/pixel, and different block size (4x4, 8x8 and, 16x16 pixels). The resulted encoding time (ET) values achieved by the proposed method were 41.53, 39.06, and 38.16 sec, respectively, for boat , butterfly, and house images of block size 4x4 pixel. These values were compared with those obtained by the traditional algorithm for the same images with the same block size, which were 1073.85, 1102.66, and 1084.92 sec, respectively. The results imply that the proposed algorithm could remarkably reduce the ET of the images in comparison with the traditional algorithm.

Production of a Three-Dimensional Triangulated Irregular Network (3D TIN) for Al-Karkh University of Science, Baghdad- Iraq

Many satellite systems take cover images like QuickBird for terrain so that these images scan be used to construct 3D models likes Triangulated Irregular Network (TIN), and Digital Elevation Model (DEM). This paper presents a production of 3D TIN for Al-Karkh University of Science in Baghdad - Iraq using QuickBird image data with pixel resolution of 0.6 m. The recent generations of high-resolution satellite imaging systems open a new era of digital mapping and earth observation. It provides not only multi-spectral and high-resolution data but also the capability for stereo mapping. The result of this study is a production of 3D satellite images of the university by merging 1 m DEM with satellite image for ROI using ArcGIS package Version 10.3.

Surface Measurement of a Large Inflatable Reflector in Cryogenic Vacuum

The metrology of membrane structures, especially inflatable, curved, optical surfaces, remains challenging. Internal pressure, mechanical membrane properties, and circumferential boundary conditions imbue highly dynamic slopes to the final optic surface. Here, we present our method and experimental results for measuring a 1 m inflatable reflector’s shape response to dynamic perturbations in a thermal vacuum chamber. Our method uses phase-measuring deflectometry to track shape change in response to pressure change, thermal gradient, and controlled puncture. We use an initial measurement as a virtual null reference, allowing us to compare 500 mm of measurable aperture of the concave f/2, 1-meter diameter inflatable optic. We built a custom deflectometer that attaches to the TVAC window to make full use of its clear aperture, with kinematic references behind the test article for calibration. Our method produces 500 × 500 pixel resolution 3D surface maps with a repeatability of 150 nm RMS within a cryogenic vacuum environment (T = 140 K, P = 0.11 Pa).

Physics-based machine learning for subcellular segmentation in living cells

Segmenting subcellular structures in living cells from fluorescence microscope images is a ground truth (GT)-deficient problem. The microscopes’ three-dimensional blurring function, finite optical resolution due to light diffraction, finite pixel resolution and the complex morphological manifestations of the structures all contribute to GT-hardness. Unsupervised segmentation approaches are quite inaccurate. Therefore, manual segmentation relying on heuristics and experience remains the preferred approach. However, this process is tedious, given the countless structures present inside a single cell, and generating analytics across a large population of cells or performing advanced artificial intelligence tasks such as tracking are greatly limited. Here we bring modelling and deep learning to a nexus for solving this GT-hard problem, improving both the accuracy and speed of subcellular segmentation. We introduce a simulation-supervision approach empowered by physics-based GT, which presents two advantages. First, the physics-based GT resolves the GT-hardness. Second, computational modelling of all the relevant physical aspects assists the deep learning models in learning to compensate, to a great extent, for the limitations of physics and the instrument. We show extensive results on the segmentation of small vesicles and mitochondria in diverse and independent living- and fixed-cell datasets. We demonstrate the adaptability of the approach across diverse microscopes through transfer learning, and illustrate biologically relevant applications of automated analytics and motion analysis.

Architecture Exploration of a Backprojection Algorithm for Real-Time Video SAR

This paper explores novel architectures for fast backprojection based video synthetic aperture radar (BP-VISAR) with multiple GPUs. The video SAR frame rate is analyzed for non-overlapped and overlapped aperture modes. For the parallelization of the backprojection process, a processing data unit is defined as the phase history data or range profile data from partial synthetic-apertures divided from the full resolution target data. Considering whether full-aperture processing is performed and range compression or backprojection are parallelized on a GPU basis, we propose six distinct architectures, each having a single-stream pipeline with a single GPU. The performance of these architectures is evaluated in both non-overlapped and overlapped modes. The efficiency of the BP-VISAR architecture with sub-aperture processing in the overlapped mode is accelerated further by filling the processing gap from the idling GPU resources with multi-stream based backprojection on multiple GPUs. The frame rate of the proposed BP-VISAR architecture with sub-aperture processing is scalable with the number of GPU devices for large pixel resolution. It can generate 4096 × 4096 video SAR frames of 0.5 m cross-range resolution in 23.0 Hz on a single GPU and 73.5 Hz on quad GPUs.

Performance-efficient distributed transfer and transformation of big spatial histopathology datasets in the cloud

Whole Slide Image (WSI) datasets are giga-pixel resolution, unstructured histopathology datasets that consist of extremely big files (each can be as large as multiple GBs in compressed format). These datasets have utility in a wide range of diagnostic and investigative pathology applications. However, the datasets present unique challenges: The size of the files, propriety data formats, and lack of efficient parallel data access libraries limit the scalability of these applications. Commercial clouds provide dynamic, cost-effective, scalable infrastructure to process these datasets, however, we lack the tools and algorithms that will transfer/transform them onto the cloud seamlessly, providing faster speeds and scalable formats. In this study, we present novel algorithms that transfer these datasets onto the cloud while at the same time transforming them into symmetric scalable formats. Our algorithms use intelligent file size distribution, and pipelining transfer and transformation tasks without introducing extra overhead to the underlying system. The algorithms, tested in the Amazon Web Services (AWS) cloud, outperform the widely used transfer tools and algorithms, and also outperform our previous work. The data access to the transformed datasets provides better performance compared to the related work. The transformed symmetric datasets are fed into three different analytics applications: a distributed implementation of a content-based image retrieval (CBIR) application for prostate carcinoma datasets, a deep convolutional neural network application for classification of breast cancer datasets, and to show that the algorithms can work with any spatial dataset, a Canny Edge Detection application on satellite image datasets. Although different in nature, all of the applications can easily work with our new symmetric data format and performance results show near-linear speed-ups as the number of processors increases.

Glass Substrate Dust Removal Using 233 fs Laser-Generated Shockwave

Eliminating dust is gaining importance as a critical requirement in the display panel manufacturing process. The pixel resolution of display panels is increasing rapidly, which means that even small dust particles on the order of a few micrometers can affect them. Conventional surface cleaning methods such as ultrasonic cleaning (USC), CO2 cleaning, and wet cleaning may not be sufficiently efficient, economical, or environment friendly. In this study, a laser shockwave cleaning (LSC) method with a 233 fs pulsed laser was developed, which is different from the laser ablation cleaning method. To minimize thermal damage to the glass substrate, the effect of the number of pulses and the gap distance between the focused laser beam and the glass substrate were studied. The optimum number of pulses and gap distance to prevent damage to the glass substrate was inferred as 500 and 20 μm, respectively. With the optimal pulse number and gap distance, cleaning efficiency was tested at a 95% removal ratio regardless of the density of the particles. The effective cleaning area was measured using the removal ratio map and compared with the theoretical value.

Assessing Wildfire Exposure to Communities and Protected Areas in Portugal

During the 2017 wildfire season in Portugal, unprecedented episodes burned 6% of the country’s area and underscored the need for a long-term comprehensive solution to mitigate future wildfire disasters. In this study, we built and calibrated a national-scale fire simulation system including the underlying fuels and weather data and used the system to quantify wildfire exposure to communities and natural areas. We simulated 10,000 fire season replicates under extreme weather to generate 1.6 million large wildfire perimeters and estimate annual burn probability and fire intensity at 100 m pixel resolution. These outputs were used to estimate wildfire exposure to buildings and natural areas. The results showed a fire exposure of 10,394 structures per year and that 30% of communities accounted for 82% of the total. The predicted burned area in natural sites was 18,257 ha yr−1, of which 9.8% was protected land where fuel management is not permitted. The main burn probability hotspots were in central and northern regions. We highlighted vital priorities to safeguard the most vulnerable communities and promote landscape management programs at the national level. The results can be useful to inform Portugal’s new national plan under implementation, where decision-making is based on a probabilistic methodology. The core strategies include protecting people and infrastructure and wildfire management. Finally, we discuss the next steps necessary to improve and operationalize the framework developed here. The wildfire simulation modeling approach presented in this study is extensible to other fire-prone Mediterranean regions where predicting catastrophic fires can help anticipate future disasters.

Evaluation of satellite retrievals of liquid clouds from the GOES-13 imager and MODIS over the midlatitude North Atlantic during the NAAMES campaign

Satellite retrievals of cloud droplet effective radius (re) and optical depth (τ) from the Thirteenth Geostationary Operational Environmental Satellite (GOES-13) and the Moderate Resolution Imaging Spectroradiometer (MODIS) aboard Aqua and Terra, based on the Clouds and the Earth's Radiant Energy System (CERES) project algorithms, are evaluated with airborne data collected over the midlatitude boundary layer during the North Atlantic Aerosols and Marine Ecosystems Study (NAAMES). The airborne dataset comprises in situ re from the Cloud Droplet Probe (CDP) and remotely sensed re and τ from the airborne Research Scanning Polarimeter (RSP). GOES-13 and MODIS (Aqua and Terra) re values are systematically greater than those from the CDP and RSP by at least 4.8 (GOES-13) and 1.7 µm (MODIS) despite relatively high linear correlation coefficients (r=0.52–0.68). In contrast, the satellite τ underestimates its RSP counterpart by −3.0, with r=0.76–0.77. Overall, MODIS yields better agreement with airborne data than GOES-13, with biases consistent with those reported for subtropical stratocumulus clouds. While the negative bias in satellite τ is mostly due to the retrievals having been collected in highly heterogeneous cloud scenes, the causes for the positive bias in satellite re, especially for GOES-13, are more complex. Although the high viewing zenith angle (∼65∘) and coarser pixel resolution for GOES-13 could explain a re bias of at least 0.7 µm, the higher GOES-13 re bias relative to that from MODIS is likely rooted in other factors. In this regard, a near-monotonic increase was also observed in GOES-13 re up to 1.0 µm with the satellite scattering angle (Θ) over the angular range 116–165∘; that is, re increases toward the backscattering direction. Understanding the variations of re with Θ will require the combined use of theoretical computations along with intercomparisons of satellite retrievals derived from sensors with dissimilar viewing geometry.