Large Computer Research Articles

Direction-of-Arrival Estimation via Sparse Bayesian Learning Exploiting Hierarchical Priors with Low Complexity.

For direction-of-arrival (DOA) estimation problems in a sparse domain, sparse Bayesian learning (SBL) is highly favored by researchers owing to its excellent estimation performance. However, traditional SBL-based methods always assign Gaussian priors to parameters to be solved, leading to moderate sparse signal recovery (SSR) effects. The reason is Gaussian priors play a similar role to l2 regularization in sparsity constraint. Therefore, numerous methods are developed by adopting hierarchical priors that are used to perform better than Gaussian priors. However, these methods are in straitened circumstances when multiple measurement vector (MMV) data are adopted. On this basis, a block-sparse SBL method (named BSBL) is developed to handle DOA estimation problems in MMV models. The novelty of BSBL is the combination of hierarchical priors and block-sparse model originating from MMV data. Therefore, on the one hand, BSBL transfers the MMV model to a block-sparse model by vectorization so that Bayesian learning is directly performed, regardless of the prior independent assumption of different measurement vectors and the inconvenience caused by the solution of matrix form. On the other hand, BSBL inherited the advantage of hierarchical priors for better SSR ability. Despite the benefit, BSBL still has the disadvantage of relatively large computation complexity caused by high dimensional matrix operations. In view of this, two operations are implemented for low complexity. One is reducing the matrix dimension of BSBL by approximation, generating a method named BSBL-APPR, and the other is embedding the generalized approximate message passing (GAMB) technique into BSBL so as to decompose matrix operations into vector or scale operations, named BSBL-GAMP. Moreover, BSBL is able to suppress temporal correlation and handle wideband sources easily. Extensive simulation results are presented to prove the superiority of BSBL over other state-of-the-art algorithms.

High efficiency independent modulation at dual-wavelength based on Pancharatnam–Berry and propagation phases

Metasurfaces capable of controlling multiple wavelengths independently have attracted broad interests these years due to their significance in multi-channel information processing applications. Previous solving strategies include spatial multiplexing or extensive searching for appropriate structures, both of which have their own disadvantageous, such as low efficiency, large computer resource requirement, or time consumption. In this paper, by combining the Pancharatnam–Berry (PB) phase and propagation phase, we propose a strategy to simplify the design complexity in a dual-wavelength metasurface system, in which two simple rectangular-shaped dielectric pillars (T1 and T2) with different aspect ratios are chosen as basic structures and crossed at the geometric center to achieve manipulation. The larger pillar T2 controls the longer wavelength through the PB phase while the smaller T1 acts as a perturbation to T2. The crossed T1&T2 is studied as a whole to tune the short wavelength. The investigations by the multipole expansion method reveal that the polarization conversion ratio of the meta-atoms is dependent on the interference of the formed multipoles. To validate the proposed strategy, a dual-wavelength achromatic metalens and a wavelength-multiplexed holographic metasurface operating at the infrared thermal imaging band are designed. Our design strategy can find widespread applications in metasurfaces where multiple objectives are required to be realized.

Estimating the carbon footprint of computational fluid dynamics

Computational resources have grown exponentially in the past few decades. These machines make possible research and design in fields as diverse as medicine, astronomy, and engineering. Despite ever-increasing computational capabilities, direct simulation of complex systems has remained challenging owing to the degrees of freedom involved. At the cusp of exascale computing, high-resolution simulation of practical problems with minimal model assumptions may soon experience a renaissance. However, growing reliance on modern computers comes at the cost of a growing carbon footprint. To illustrate this, we examine historic computations in fluid dynamics where larger computers have afforded the opportunity to simulate flows at increasingly relevant Reynolds numbers. Under a variety of flow configurations, the carbon footprint of such simulations is found to scale roughly with the fourth power of Reynolds number. This is primarily explained by the computation cost in core-hours, which is also described by similar scaling, though regional differences in renewable energy use also play a role. Using the established correlation, we examine a large database of simulations to develop estimates for the carbon footprint of computational fluid dynamics in a given year. Collectively, the analysis provides an additional benchmark for new computations where, in addition to balancing considerations of model fidelity, carbon footprint should also be considered.

Data Science Has Exploded Across the Industry Over the Past 75 Years

In 1957, the Journal of Petroleum Technology published an article titled “Application of Large Computers to Reservoir Engineering Problems.” That was the first reference in JPT to what became known as supercomputers. The high-speed computers used in the 1957 article were capable of performing 60 million operations in about 3½ hours. They were proposed to analyze the thorny problem of multiphase flow. “It presently appears that the large computer will be required to investigate multiphase flow and to predict the flow behavior of oil and gas reservoirs considering two- and three-space dimensions.” Now, 67 years later, supercomputer speeds are measured in trillions of operations per second, and investigating multiphase flow is just one of the many uses. This extreme growth in speed has given rise to artificial intelligence (AI) and machine learning (ML), which has found its way into almost every corner of the petroleum industry. The Society of Petroleum Engineers and JPT has kept up with the digital advances in the industry, holding countless conferences, symposia, and other meetings centered on the digital aspects of the industry and recently creating the Data Science and Engineering Analytics (DSEA) technical discipline. In 2019, SPE launched the Data Science and Digital Engineering in Upstream Oil and Gas online publication. Four leaders in the oil and gas digital space shared their views on the current state of data science in the industry and the future of the discipline. Sushma Bhan, currently on the board of directors for Ikon Science, is SPE’s Technical Director for DSEA. Before joining Ikon, she worked for Shell for 32 years, eventually rising to the role of chief data officer for subsurface and wells. “My journey in the oil and gas industry began in 1988, when I joined Shell’s Production Computing Assisted Operations team as a programmer analyst,” Bhan said. “During my time there, I gained valuable exposure to field production operations and developed a deep understanding of real‑time data.” Jim Crompton, who claims he is mostly retired, is an associate professor of petroleum engineering at the Colorado School of Mines, a faculty fellow at the school’s Payne Institute for Public Policy, and director of Reflections Data Consulting. “My academic career started out in exploration geophysics, so my first introduction to engineering analytics came from processing seismic data at Chevron Geophysical Company in Houston,” Crompton said. “I studied earthquake seismology in graduate school, but, when I graduated, I learned the oil and gas industry paid a lot more than the USGS did, so my career goals changed.” Shahab D. Mohaghegh is a professor at West Virginia University and the president of Intelligent Solutions. “Since I started working for the petroleum industry throughout the world using artificial intelligence in 2000, I was able to work with actual data (field measurements) that has been saved by all the companies,” Mohaghegh said. “Working with actual data using AI provided several technologies that have been incredibly fantastic compared to what we have done in the past.” Pushpesh Sharma, the chair of the DSEA Technical Section, holds a PhD degree in chemical engineering from the University of Houston and is senior product manager for Aspen Technology. “Till a few years ago, the focus was on proving the efficacy of data science/machine learning methods for energy use cases,” Sharma said. “However, in recent years, the concerns are around large-scale deployment, maintenance, and the black‑box nature of ML models. Because of that, I started seeing increased focus on the deployment, explainability, and trust of ML models.”

A numerical method for ergodic optimal control of switching diffusions with reflection

The ergodic or long run average cost control problem for diffusions is one of the classical problems of stochastic control that still eludes a completely satisfactory treatment. This is certainly true for the setting in which the systems to be controlled is modeled by the solution of a switching stochastic differential equation with reflection (SSDER). In this paper, we advance in this rather difficult problem by setting forth preliminary results for the unidimensional case. Besides carving out a numerical method, our treatment of the ergodic control in this scenario straddles issues of existence and uniqueness of solution of the SSDER and a verification theorem for the associated HJB equation. We conclude by illustrating the effectiveness of the method considering the control of energy consumption in a large parallel processing computer system composed of one queue and several processing stations.

GMiner++: Boosting GPU-based frequent itemset mining by reducing redundant computations

Frequent itemset mining (FIM) is increasingly important in fundamental data mining techniques. However, the applicability of existing FIM methods is limited, mainly because of their performance. The expected performance improvement is limited owing to the exploitation of only a single thread, despite numerous efficient single-threaded FIM methods being proposed. Numerous parallel FIM methods have been devised using graphic processing units (GPU) or multicore central processing units (CPUs) to overcome the shortcomings of these methods. However, when extracting patterns from large amounts of data, multi-threaded FIM methods exhibit a similar performance tendency to single-threaded FIM methods, because of their large memory footprints and computations. Hence, we propose GMiner++, a memory-efficient GPU-based FIM method equipped with several GPUs. We propose a sub-database of the same size called bit array blocks, which contains pre-calculated bit arrays of F1∪P(IK). These bit arrays are repeatedly exploited during mining tasks using an elegant probabilistic model. GMiner++ can obtain frequent patterns and use several GPUs only by using the bit array blocks and the occurrence update scheme. The proposed method decreased redundant computations using pre-calculated bit arrays with bit array blocks. In addition, GMiner++ does not create intermediate data during mining tasks to increase robustness and reduce memory footprints. Simulation results demonstrate that GMiner++ outperformed existing FIM methods concerning performance and scalability with increasing robustness.

Sustainability of Data Center Digital Twins with Reinforcement Learning

The rapid growth of machine learning (ML) has led to an increased demand for computational power, resulting in larger data centers (DCs) and higher energy consumption. To address this issue and reduce carbon emissions, intelligent design and control of DC components such as IT servers, cabinets, HVAC cooling, flexible load shifting, and battery energy storage are essential. However, the complexity of designing and controlling them in tandem presents a significant challenge. While some individual components like CFD-based design and Reinforcement Learning (RL) based HVAC control have been researched, there's a gap in the holistic design and optimization covering all elements simultaneously. To tackle this, we've developed DCRL-Green, a multi-agent RL environment that empowers the ML community to design data centers and research, develop, and refine RL controllers for carbon footprint reduction in DCs. It is a flexible, modular, scalable, and configurable platform that can handle large High Performance Computing (HPC) clusters. Furthermore, in its default setup, DCRL-Green provides a benchmark for evaluating single as well as multi-agent RL algorithms. It easily allows users to subclass the default implementations and design their own control approaches, encouraging community development for sustainable data centers. Open Source Link: https://github.com/HewlettPackard/dc-rl

Practical Sentiment Analysis for Education: The Power of Student Crowdsourcing

Sentiment analysis provides a promising tool to automatically assess the emotions voiced in written student feedback such as periodically collected unit-of-study reflections. The commonly used dictionary-based approaches are limited to major languages and fail to capture contextual differences. Pretrained large language models have been shown to be biased and online versions raise privacy concerns. Hence, we resort to traditional supervised machine learning (ML) approaches which are designed to overcome these issues by learning from domain-specific labeled data. However, these labels are hard to come by -- in our case manually annotating student feedback is prone to bias and time-consuming, especially in high-enrollment courses. In this work, we investigate the use of student crowdsourced labels for supervised sentiment analysis for education. Specifically, we compare crowdsourced and student self-reported labels with human expert annotations and use them in various ML approaches to evaluate the performance on predicting emotions of written student feedback collected from large computer science classes. We find that the random forest model trained with student-crowdsourced labels tremendously improves the identification of reflections with negative sentiment. In addition to our quantitative study, we describe our crowdsourcing experiment which was intentionally designed to be an educational activity in an introduction to data science course.

AQ-DETR: Low-Bit Quantized Detection Transformer with Auxiliary Queries

DEtection TRansformer (DETR)-based models have achieved remarkable performance. However, they are accompanied by a large computation overhead cost, which significantly prevents their applications on resource-limited devices. Prior arts attempt to reduce the computational burden of DETR using low-bit quantization, while these methods sacrifice a severe significant performance on weight-activation-attention low-bit quantization. We observe that the number of matching queries and positive samples affect much on the representation capacity of queries in DETR, while quantifying queries of DETR further reduces its representational capacity, thus leading to a severe performance drop. We introduce a new quantization strategy based on Auxiliary Queries for DETR (AQ-DETR), aiming to enhance the capacity of quantized queries. In addition, a layer-by-layer distillation is proposed to reduce the quantization error between quantized attention and full-precision counterpart. Through our extensive experiments on large-scale open datasets, the performance of the 4-bit quantization of DETR and Deformable DETR models is comparable to full-precision counterparts.

Tensor networks for solving the time-independent Boltzmann neutron transport equation

Tensor network techniques, known for their low-rank approximation ability that breaks the curse of dimensionality, are emerging as a foundation of new mathematical methods for ultra-fast numerical solutions of high-dimensional Partial Differential Equations (PDEs). Here, we present a mixed Tensor Train (TT)/Quantized Tensor Train (QTT) approach for the numerical solution of time-independent Boltzmann Neutron Transport equations (BNTEs) in Cartesian geometry. Discretizing a realistic three-dimensional (3D) BNTE by (i) diamond differencing in space, (ii) multigroup-in-energy, and (iii) discrete ordinates collocation in angle leads to large generalized eigenvalue problems that generally require a matrix-free approach and large computer clusters. Starting from this discretization, we construct a TT representation of the PDE fields and discrete operators, followed by a QTT representation of the TT cores. We then solve the tensorized generalized eigenvalue problem using a fixed-point scheme with tensor network optimization techniques. We validate our approach by applying the method to two examples of 3D neutron transport problems, currently solved by the Los Alamos National Laboratory PARallel TIme-dependent SN (PARTISN) solver.1 We demonstrate that our TT/QTT method, executed on a standard desktop computer, leads to large compression. This allows for the storage of terrabyte-sized neutron angular flux eigenvectors in megabytes. Additionally, we create megabyte-sized full access TT representations of yottabyte-sized transport matrix operators. By leveraging the TT operators and solution methods, we obtain a 7500 times speedup when compared to the PARTISN solution time with an error of less than 10−5.

Learning model combining convolutional deep neural network with a self-attention mechanism for AC optimal power flow

Alternating current optimal power flow (OPF) analysis is critical for efficient and reliable operation of power systems. For large systems or repetitive computations, the traditional methods such as the direct and gradient methods, or non-traditional methods, such as the genetic algorithm and simulating annealing, are time-consuming and unsuitable for real-time computing. The work in this paper proposes a novel framework to obtain the optimal solution of power flow in real-time using a combination of convolutional neural networks and a self-attention mechanism. All parameters of the power networks are rearranged in an image-like shape of a multi-channel image where each channel is a two-dimensional matrix. The proposed approach is adaptive with every input size of power systems as well as frequent variations of network topologies without intervention to the framework core. The encompassment of all power system contexts in which all parameters of internal elements, generation costs, and topology information are included, contributes to the higher accuracy of inference compared to other current machine-learning-based OPF-solving methods. Besides, the proposed framework established on ubiquitous platforms is effortlessly integrated into current infrastructures of power systems, and the great efficiency along with the computation speed may serve as a critical point for practical implications, such as enabling faster decision-making during real-time operations, predicting system contingencies, and remedial actions based on an offline pre-trained model. This supervised learning process is applied to the dataset of four case studies of meshed power systems: the IEEE 5-bus system (IEEE-5), the IEEE 30-bus system (IEEE-30), the IEEE 39-bus system (IEEE-39), and the IEEE 57-bus system (IEEE-57) to prove the efficacy of the proposed method.

SPASTC: a Spatial Partitioning Algorithm for Scalable Travel-time Computation

Travel-time computation with large transportation networks is often computationally intensive for two main reasons: 1) large computer memory is required to handle large networks; and 2) calculating shortest-distance paths over large networks is computing intensive. Therefore, previous research tends to limit their spatial extent to reduce computational intensity or resolve computational intensity with advanced cyberinfrastructure. In this context, this article describes a new Spatial Partitioning Algorithm for Scalable Travel-time Computation (SPASTC) that is designed based on spatial domain decomposition with computer memory limit explicitly considered. SPASTC preserves spatial relationships required for travel-time computation and respects a user-specified memory limit, which allows efficient and large-scale travel-time computation within the given memory limit. We demonstrate SPASTC by computing spatial accessibility to hospital beds across the conterminous United States. Our case study shows that SPASTC achieves significant efficiency and scalability making the travel-time computation tens of times faster.

Spatiotemporal correlation based self-adaptive pose estimation in complex scenes

In the current digital era, large-scale Artificial Intelligence (AI) models have fundamentally changed the landscape of the AI field. These models excel at extracting complex patterns and information from massive datasets, making them indispensable in a wide range of domains. With the advancement of large-scale AI models, pose estimation has achieved improved performance in managing large data sets and complex computations. However, most existing pose estimation methods suffer from the problem of inaccurate pose graph extraction in complex scenes. Moreover, the existing methods all focus on the message, ignoring the fact that it may not be optimal for the graph structure. This fixed graph structure is difficult to reflect the real dependency relationship between joint points. This study proposes a Character Redetection Model (CRM) to dynamically adjust the foreground extraction strategies to solve these problems. The CRM locates the target position based on the rich spatiotemporal information between the adjacent frames, effectively solving the problem of inaccurate foreground acquisition caused by occlusion and blur in complex scenes. In addition, an Uncertainty Graph Structure (UGS) is designed, which adaptively learns the topological structure of graphs with different Graph Convolutional Network (GCN) layers and skeleton samples in an end-to-end manner and mines the hidden space dependency between data to accurately model the different poses of the characters. The experimental results show that the proposed method achieves a superior performance compared to that of the State-Of-The-Art (SOTA) method, with a percentage of correct keypoints (PCK) value of 94.6% at the Max Planck Institute for Informatics (MPII), along with the mean Average Precision (mAP) values of 83.3% and 82.1% on the PoseTrack2017 and classroom pose datasets, respectively.

Video gaming industry in the US

Generations after generations, the entertainment industry evolved multiple folds, be it music, be it movies, be it sports, be it video games. Each industry doesn’t only set trends each decade but also changes peoples’ lifestyle. Video games are something which is involved almost in every person’s life. Some people play them when they’re young, some play in their teens, some play after they come home from office and some play after they retire. The video games were simply designed simulations that was started in 1950’s and 1960’s by computer scientists on the early computers. In 1952, British professor A.S. Douglas created OXO, also known as noughts and crosses or tic-tac-toe, as part of his doctoral dissertation at the University of Cambridge. In 1958, physicist William Higinbotham developed “Tennis for Two” on a large analog computer and connected oscilloscope screen. In 1962, Steve Russell at the Massachusetts Institute of Technology (MIT) invented Spacewar! a computer-based space combat video game. It ran on the Programmed Data Processor-1 (PDP-1), a cutting-edge computer primarily found at universities. Spacewar! was the first video game that could be played on multiple computer installations. In 1967, Ralph Baer and his team at Sanders Associates, Inc., created a prototype multiplayer, multi-program video game system that could be played on a television. Known as “The Brown Box,” this invention led to the release of the Magnavox Odyssey in 1972, the first video game home console. In 1977, Atari launched the Atari 2600 (also known as the Video Computer System). This home console featured joysticks and interchangeable game cartridges, marking the beginning of the second generation of video game consoles. In 2023, the gaming industry reportedly earns more than film industry and music industry combined which is around $249.60 billion. According to a report by SuperData Research, the global gaming market was valued at $159.3 billion in 2020 and is estimated to reach $268.8 billion by 2025. This is already over three times the revenue of music industry and film industry combined. [2] The 2024 global video game market's projected value of $179.1 billion, with a healthy 6.4% CAGR, bodes well for the US industry. As a key player, the US drives this growth through successful titles like Call of Duty and Grand Theft Auto. A mobile gaming market experiencing significant growth and innovation, led by companies like Zynga, strengthens its longstanding dominance in the international arena. This success attracts international investment and talent, promoting innovation and pushing the industry forward.

Retrieval of chromophore concentration changes in a digital human head model using analytical mean partial pathlengths of photons.

Continuous-wave functional near-infrared spectroscopy has proved to be a valuable tool for assessing hemodynamic activity in the human brain in a non-invasively and inexpensive way. However, most of the current processing/analysis methods assume the head is a homogeneous medium, and hence do not appropriately correct for the signal coming from the scalp. This effect can be reduced by considering light propagation in a layered model of the human head, being the Monte Carlo (MC) simulations the gold standard to this end. However, this implies large computation times and demanding hardware capabilities. In this work, we study the feasibility of replacing the homogeneous model and the MC simulations by means of analytical multilayered models, combining in this way, the speed and simplicity of implementation of the former with the robustness and accuracy of the latter. Oxy- and deoxyhemoglobin (HbO and HbR, respectively) concentration changes were proposed in two different layers of a magnetic resonance imaging (MRI)-based meshed model of the human head, and then these changes were retrieved by means of (i)a typical homogeneous reconstruction and (ii)a theoretical layered reconstruction. Results suggest that the use of analytical models of light propagation in layered models outperforms the results obtained using traditional homogeneous reconstruction algorithms, providing much more accurate results for both, the extra- and the cerebral tissues. We also compare the analytical layered reconstruction with MC-based reconstructions, achieving similar degrees of accuracy, especially in the gray matter layer, but much faster (between 4 and 5 orders of magnitude). We have successfully developed, implemented, and validated a method for retrieving chromophore concentration changes in the human brain, combining the simplicity and speed of the traditional homogeneous reconstruction algorithms with robustness and accuracy much more similar to those provided by MC simulations.

On the Swelling of Polymer Network Strands.

Large scale computer simulations are employed to analyze the conformations of network strands in polymer networks at preparation conditions (characterized by a polymer volume fraction of ϕ0 ) and when swollen to equilibrium (characterized by a polymer volume fraction ϕ < ϕ0 ). Network strands in end-linked model networks are weakly stretched and partially swollen at preparation conditions as compared to linear polymers in the same solvent at ϕ0 . Equilibrium swelling causes non-ideal chain conformations characterized by an effective scaling exponent approaching 7/10 on intermediate length scales for increasing overlap of the chains. The chain size in a network consists of a fluctuating and a time average "elastic" contribution. The elastic contribution swells essentially affinely ∝(ϕ0 /ϕ)2/3 , whereas the swelling of the fluctuating part lies between the expected swelling of the entanglement constraints and the swelling of non-cross-linked chains in a comparable semi-dilute solution. The total swelling of chain size results from the changes of both fluctuating and non-fluctuatingcontributions.

Modeling of an Irrigation Main Canal Pool based on a [formula omitted]-[formula omitted] System Identification

Dynamic models of main irrigation canals are necessary to carry out real-time canal flow control and supervision. Since the hydraulic equations of canal dynamics are nonlinear, their solution involves large computations that impede their use in real-time applications. Then these tasks are often implemented using simple local linear models obtained around an operating point, that are unable to capture the dynamics when large canal operation changes happen.In this paper, we propose a nonlinear dynamic model for main irrigation canal pools based on a new Artificial Neural Network (ANN). This ANN outperforms the above local models because it includes a new input which is the downstream water level of the previous pool. This new architecture with this new input yields models that accurately reproduce the behavior of the canal in a wide range of operations while being simple enough to allow real-time control and supervision. The high accuracy of the proposed model as regards reproducing the observed data is validated for different flow regimes based on experiments developed in a laboratory prototype of the hydraulic canal and in a real main irrigation canal.The contribution of this work is such a general and accurate but simple model of the dynamics of a canal pool. Its novelty is not only being able to reproduce the behavior at different operating points of the canal but also being able to reproduce, in a single model, the free and submerged flow regimes — which are described by dynamic equations with different structures. Moreover, we implement a new classification model to recognize the flow regimes of the irrigation canals using pattern recognition.

Tomography-based digital twin of Nd-Fe-B permanent magnets

Many functional materials have been designed at the multiscale level. To properly simulate their physical properties, large and sophisticated computer models that can replicate microstructural features with nanometer-scale accuracy are required. This is the case for permanent magnets, which exhibit a long-standing problem of a significant offset between the simulated and experimental coercivities. To overcome this problem and resolve the Brown paradox, we propose an approach to construct large-scale finite element models based on the tomographic data from scanning electron microscopy. Our approach reconstructs a polycrystalline microstructure with actual shape, size, and packing of the grains as well as the individual regions of thin intergranular phase separated by triple junctions. Such a micromagnetic model can reproduce the experimental coercivity of ultrafine-grained Nd-Fe-B magnets along with its mechanism according to the angular dependence of coercivity. Furthermore, a remarkable role of thin triple junctions as nucleation centers for magnetization reversal is revealed. The developed digital twins of Nd-Fe-B permanent magnets can assist their optimization toward the ultimate coercivity, while the proposed tomography-based approach can be applied to a wide range of polycrystalline materials.

Model validation of hollow embossing rolling for bipolar plate forming

Hollow embossing rolling is a promising forming technology for metallic bipolar plates because of the high achievable production rates. However, the simulation-based process optimization is impeded by the incremental forming character and modeling of fine channel structures, which leads to large model sizes and long computation times. This paper presents a shell-based finite element approach validated by experimental forming tests using a miniaturized test geometry with typical discontinuities and varying channel orientations. The rolling experiments demonstrated that implementing restraining tension effectively decreases wrinkling, allowing successful forming of the selected test geometry by hollow embossing rolling. It was found that representing the manufacturing-related decreased rolling gap combined with the rolling gap changes due to roll system elasticity in the numerical model is essential for model accuracy. An optimized model approach with spring-controlled rollers was developed, which considers the effect of load-dependent rolling gap changes. With this approach the applied model achieves sufficient model accuracy for technological process simulation and optimization.

DISH: Digital image steganography using stochastic-computing with high-capacity

AbstractStochastic computing is a relatively new approach to computing that has gained interest in recent years due to its potential for low-power and high-noise environments. It is a method of computing that uses probability to represent and manipulate data, therefore it has applications in areas such as signal processing, machine learning, and cryptography. Stochastic steganography involves hiding a message within a cover image using a statistical model. Unlike traditional steganography techniques that use deterministic algorithms to embed the message, stochastic steganography uses a probabilistic approach to hide the message in a way that makes it difficult for an adversary to detect. Due to this error robustness and large bit streams stochastic computing, they are well suited for high capacity and secure image steganography. In this paper, as per the authors’ best knowledge, image steganography using stochastic computing based on linear feedback shift register (LFSR) is proposed for the first time. In the proposed technique, the cover image is converted to stochastic representation instead of the binary one, and then a secret image is embedded in it. The resulting stego image has a high PSNR value transmitted with no visual trace of the hidden image. The final results are stego image with PSNR starting from 30 dB and a maximum payload up to 40 bits per pixel (bpp) with an effective payload up to 28 bpp. The proposed method achieves high security and high capability of the number of stored bits in each pixel. Thus, the proposed method can prove a vital solution for high capacity and secure image steganography, which can then be extended to other types of steganography.