Large-scale Machine Research Articles

Vision Guided Dynamic Synchronous Path Tracking Control of Dual Manipulator Cooperative System

Abstract Compared with a single manipulator manufacturing cell, a dual manipulator cooperative system has more advantages in reconfigurability and flexibility. However, there are calibration errors and multi-source disturbances in the collaborative process, which lead to the processing trajectory accuracy defects of large-scale associated machining features. To solve the above problems, a practical path tracking synchronous control algorithm based on position-based visual servoing (PBVS) is proposed in this paper for the dual manipulator cooperative system, the proposed dynamic path tracking cross-coupled sliding mode controller (PTCSMC) scheme can realize dynamic paths correction while executing the pre-planned paths. In addition, for the cross-coupled technology to be applied into the proposed control algorithm for dynamic path tracking based on the real-time feedback of the highly repeatable 3D visual measurement instrument (VMI), the tracking and synchronous errors of the dual manipulators converge synchronously to zero. Finally, the stability of proposed control algorithm is proven by the Lyapunov method. In the end, the real-time line and circle path tracking experimental results using two industrial manipulators demonstrate the effectiveness of the proposed algorithm.

Multinomial classification of NLRP3 inhibitory compounds based on large scale machine learning approaches.

The role of NLRP3 inflammasome in innate immunity is newly recognized. The NLRP3 protein is a family of nucleotide-binding and oligomerization domain-like receptors as well as a pyrin domain-containing protein. It has been shown that NLRP3 may contribute to the development and progression of a variety of diseases, such as multiple sclerosis, metabolic disorders, inflammatory bowel disease, and other auto-immune and auto-inflammatory conditions. The use of machine learning methods in pharmaceutical research has been widespread for several decades. An important objective of this study is to apply machine learning approaches for the multinomial classification of NLRP3 inhibitors. However, data imbalances can affect machine learning. Therefore, a synthetic minority oversampling technique (SMOTE) has been developed to increase the sensitivity of classifiers to minority groups. The QSAR modelling was performed using 154 molecules retrieved from the ChEMBL database (version 29). The accuracy of the multiclass classification top six models was found to fall within ranges of 0.99 to 0.86, and log loss ranges of 0.2 to 2.3, respectively. The results showed that the receiver operating characteristic curve (ROC) plot values significantly improved when tuning parameters were adjusted and imbalanced data was handled. Moreover, the results demonstrated that SMOTE offers a significant advantage in handling imbalanced datasets as well as substantial improvements in overall accuracy of machine learning models. The top models were then used to predict data from unseen datasets. In summary, these QSAR classification models exhibited robust statistical results and were interpretable, which strongly supported their use for rapid screening of NLRP3 inhibitors.

Persistent Homology through Image Segmentation (Student Abstract)

The efficacy of topological data analysis (TDA) has been demonstrated in many different machine learning pipelines, particularly those in which structural characteristics of data are highly relevant. However, TDA's usability in large scale machine learning applications is hindered by the significant computational cost of generating persistence diagrams. In this work, a method that allows this computationally expensive process to be approximated by deep neural networks is proposed. Moreover, the method's practicality in estimating 0-dimensional persistence diagrams across a diverse range of images is shown.

Hydropower Planning in Combination with Batteries and Solar Energy

Battery storage is an important factor for power systems made up of renewable energy sources. Technologies for battery storage are crucial to accelerating the transition from fossil fuels to renewable energy. Between responding to electricity demand and using renewable energy sources, battery storage devices will become increasingly important. The aim of this study is to examine how battery storage affects a power system consisting of solar and hydroelectric energy and to draw conclusions about whether energy storage recommends a power system. The method involves designing a model of eight real cascade hydropower power plants and solving an optimization problem. This power system model is based on existing hydroelectric power plants powered by solar energy and batteries in the Turkish cities of Yozgat and Tokat. A case study with four different battery capacities in the system was carried out to assess the implications of energy storage in the power system. The stochastic nonlinear optimization problem was modeled for 72 h and solved with the MATLAB programming tool. The stochastic Quasi-Newton method performs very well in hybrid renewable problems arising from large-scale machine learning. When solar energy and batteries were added to the system, the maximum installed wind power was found to be 2 MW and 3.6 MW, respectively. In terms of profit and hydropower planning, a medium-proportion battery was found to be the most suitable. Increased variability in hydropower generation results from the installation of an energy storage system.

Active learning on a programmable photonic quantum processor

Training a quantum machine learning model generally requires a large labeled dataset, which incurs high labeling and computational costs. To reduce such costs, a selective training strategy, called active learning (AL), chooses only a subset of the original dataset to learn while maintaining the trained model’s performance. Here, we design and implement two AL-enpowered variational quantum classifiers to investigate the potential applications and effectiveness of AL in quantum machine learning. Firstly, we build a programmable free-space photonic quantum processor, which enables the programmed implementation of various hybrid quantum-classical computing algorithms. Then, we code the designed variational quantum classifier with AL into the quantum processor, and execute comparative tests for the classifiers with and without the AL strategy. The results validate the great advantage of AL in quantum machine learning, as it saves at most labeling efforts and percent computational efforts compared to the training without AL on a data classification task. Our results inspire AL’s further applications in large-scale quantum machine learning to drastically reduce training data and speed up training, underpinning the exploration of practical quantum advantages in quantum physics or real-world applications.

Communication-efficient ADMM-based distributed algorithms for sparse training

In large-scale distributed machine learning (DML), the synchronization efficiency of the distributed algorithm becomes a critical factor that affects the training time of machine learning models as the computing scale increases. To address this challenge, we propose a novel algorithm called Grouped Sparse AllReduce based on the 2D-Torus topology (2D-TGSA), which enables constant transmission traffic that does not change with the number of workers. Our experimental results demonstrate that 2D-TGSA outperforms several benchmark algorithms in terms of synchronization efficiency. Moreover, we integrate the general form consistent ADMM with 2D-TGSA to develop a distributed algorithm (2D-TGSA-ADMM) that exhibits excellent scalability and can effectively handle large-scale distributed optimization problems. Furthermore, we enhance 2D-TGSA-ADMM by adopting the resilient adaptive penalty parameter approach, resulting in a new algorithm called 2D-TGSA-TPADMM. Our experiments on training the logistic regression model with ℓ1-norm on the Tianhe-2 supercomputing platform demonstrate that our proposed algorithm can significantly reduce the synchronization time and training time compared to state-of-the-art methods.

Polarized fusion and potential in situ tests of fuel polarization survival in a tokamak plasma

The use of spin-polarized fusion fuels would provide a significant boost towards the ignition of a burning plasma. The cross section for D + T → α + n, would be increased by 1.5 if the fuels were injected with parallel polarization. Furthermore, our simulations demonstrate additional non-linear power gains in large-scale machines such as ITER, due to increased alpha heating. Such benefits require the survival of spin polarizations for periods comparable to the particle confinement time. During the 1980s, calculations predicted that polarizations could survive a plasma environment, although concerns persisted regarding the cumulative impacts of wall recycling. In that era, technical challenges prevented direct tests and left the large scale fueling of a power reactor beyond reach. Over the last decades, this situation has changed dramatically. Detailed simulations of ITER have predicted negligible wall recycling in a high-power reactor, and recent advances in laser-driven sources project the capability of producing large quantities of ∼100% polarized D and T. The remaining crucial step is an in-situ demonstration of polarization survival in a plasma. For this, we outline a measurement strategy using the isospin-mirror reaction, D + 3He → α + p. Polarized 3He avoids the complexities of handling tritium, while encompassing the same spin-physics. We evaluate two methods of delivering deuterium, using dynamically polarized Lithium-Deuteride (with vector polarization PV D of 70%) or frozen-spin Hydrogen-Deuteride (with PV D of 40%), together with a method of injecting optically-pumped 3He (with 65% polarization). Pellets of these materials all have long polarization decay times (∼6 min for LiD at 2 K, ∼2 months for HD at 2 K, and ∼3 d for 3He at 77 K), all far greater than a plasma shot in a research tokamak such as DIII-D (∼20 s). Both species can be propelled from a single cryogenic injection gun. We review plasma requirements and strategies for detecting polarization survival. Polarization alters both fusion yields and the angular distribution of fusion products, and each of these provides a potential signal. In this paper we simulate a selection of shots with similar characteristics in a future high-Tion H plasma, and find ratios of yields from shots with fuel spins parallel and antiparallel reaching 1.3 (HD + 3He) to 1.6 (LiD + 3He) over a wide range of poloidal angles. (A companion paper finds sensitivity to fusion product angular distributions as reflected in the pitch angles of protons and alphas reaching the plasma facing wall.)

Gradient Coding With Dynamic Clustering for Straggler-Tolerant Distributed Learning

Distributed implementations are crucial in speeding up large scale machine learning applications. Distributed gradient descent (GD) is widely employed to parallelize the learning task by distributing the dataset across multiple workers. A significant performance bottleneck for the per-iteration completion time in distributed synchronous GD is straggling workers. Coded distributed computation techniques have been introduced recently to mitigate stragglers and to speed up GD iterations by assigning redundant computations to workers. In this paper, we introduce a novel paradigm of dynamic coded computation, which assigns redundant data to workers to acquire the flexibility to dynamically choose from among a set of possible codes depending on the past straggling behavior. In particular, we propose gradient coding (GC) with dynamic clustering, called GC-DC, and regulate the number of stragglers in each cluster by dynamically forming the clusters at each iteration. With time-correlated straggling behavior, GC-DC adapts to the straggling behavior over time; in particular, at each iteration, GC-DC aims at distributing the stragglers across clusters as uniformly as possible based on the past straggler behavior. For both homogeneous and heterogeneous worker models, we numerically show that GC-DC provides significant improvements in the average per-iteration completion time without an increase in the communication load compared to the original GC scheme.

A kernel search algorithm for virtual machine consolidation problem in cloud computing

Virtual machine consolidation refers to the process of reallocating virtual machines across a set of target servers. It can be formulated as a mixed integer linear programming problem, which is known to be NP-hard. In this paper, we propose a kernel search (KS) heuristic algorithm based on hard variable fixing to quickly obtain high-quality solutions for large-scale virtual machine consolidation problems (VMCPs). Since existing variable fixing strategies in KS algorithms may result in some VMCP instances being infeasible, our proposed KS algorithm employs a more efficient strategy to choose a set of fixed variables based on their corresponding reduced costs. Our numerical results on VMCP instances demonstrate that our proposed KS algorithm significantly outperforms mixed integer linear programming solvers in terms of CPU time. Moreover, our proposed strategy of variable fixing improves the efficiency of the KS algorithm significantly, with negligible degradation in solution quality.

Detecting the risk of bullying victimization among adolescents: A large-scale machine learning approach

There is an increasing interest in using machine learning methods to identify risk factors for problematic behaviors. The current study tested and compared six machine learning algorithms: Logistic Regression, Naive Bayes, Decision Tree, Random Forest, K-Nearest Neighbors (KNN), and Light Gradient Boosting Machine (LightGBM), to detect risk factors for both traditional bullying victimization and cyberbullying victimization among Chinese adolescents. The Random Forest algorithm and LightGBM algorithm obtained similar accuracy and precision, and outperformed other four algorithms. We then combined the feature importance of LightGBM and Random Forest algorithms to evaluate the predictive power of 40 potentially relevant personal, educational, social and psychological factors in predicting bullying victimization, achieving better accuracy and higher performance. These results showed that the combined model can distinguish high-risk and low-risk adolescents for both types of bullying victimization based on a few easy-to-find variables. By comparing the relative significance of each factor, the current study also found mental illness, physical illness, and unhealthy living environments as having the highest values in predicting bullying victimization. Thus, the recommended model has a great application value in preventing bullying victimization among Chinese adolescents.

Experimental Study on the Effect of Coarse Grain Content on the Dilatancy and Particle Breakage Characteristics of Coarse-Grained Soils

The dilatancy deformation and particle breakage effects of coarse-grained soil have a significant influence on the safety and stability of high earth and rock-fill dams. To study the shear deformation and particle breakage characteristics of coarse-grained soil, four groups of soil sample with different coarse grain contents were to conduct consolidated drained triaxial compression tests by DJSZ-150 large-scale triaxial compression test machine. The results showed that the volume of coarse-grained soil changes significantly under deviation stress. The soil samples underwent positive dilatancy under the confining pressures of 200 kPa and 400 kPa but negative dilatancy under the confining pressures of 600 kPa and 800 kPa. When increasing the coarse grain contents in the soil sample under the same confining pressure, the positive dilatancy was more significant. A three-parameter dilatancy equation of coarse-grained soil considering confining pressure was established, in which the parameter L = 4 could serve as a criterion for whether the coarse-grained soil underwent positive dilatancy. Under the load, the coarse-grained soil showed three particle breakage modes: abrasion, attrition, and fracture. The coarse-grained soil particle size distribution before and after breakage conformed to the fractal characteristics, and the particle breakage rate increased significantly with larger soil particle size fractal dimension differences before and after the test. The coarse grain content and the confining pressure are two important factors affecting the mechanical properties of coarse-grained soil, and they control the shear deformation of soil.

Wind turbine structural control using H-infinity methods

Wind turbines are complex dynamic structures, exposed to random wind input exciting several structural modes. In recent years, the wind turbines have evolved to large-scale machines, with rotor diameter and tower height attaining unforeseen dimensions compared to initial commercial wind turbines. In this context, the structure design has become more flexible and susceptible to dynamic actions, rising the need for the development of active structural control methodologies. On the other hand, the wind turbine dispose of standard control systems, dedicated to operational control. The use of these control systems to mitigate wind turbine loads is feasible due to the dynamic couplings within the structure, and could provide a significant comparative advantage to the new wind turbine structures. In order to provide structural control from conventional wind turbine control, it is necessary to deploy a suitable modeling and the use of advanced control methods. In this paper, the use of H∞ control design is combined with wind turbine dynamics equations to develop an active structural control to reduce wind turbine tower loading. Dynamic simulations under steady and turbulent conditions demonstrate the effectiveness of the proposed approach.

Precision engineering of biological function with large-scale measurements and machine learning.

As synthetic biology expands and accelerates into real-world applications, methods for quantitatively and precisely engineering biological function become increasingly relevant. This is particularly true for applications that require programmed sensing to dynamically regulate gene expression in response to stimuli. However, few methods have been described that can engineer biological sensing with any level of quantitative precision. Here, we present two complementary methods for precision engineering of genetic sensors: in silico selection and machine-learning-enabled forward engineering. Both methods use a large-scale genotype-phenotype dataset to identify DNA sequences that encode sensors with quantitatively specified dose response. First, we show that in silico selection can be used to engineer sensors with a wide range of dose-response curves. To demonstrate in silico selection for precise, multi-objective engineering, we simultaneously tune a genetic sensor's sensitivity (EC50) and saturating output to meet quantitative specifications. In addition, we engineer sensors with inverted dose-response and specified EC50. Second, we demonstrate a machine-learning-enabled approach to predictively engineer genetic sensors with mutation combinations that are not present in the large-scale dataset. We show that the interpretable machine learning results can be combined with a biophysical model to engineer sensors with improved inverted dose-response curves.

A large-scale machine learning study of sociodemographic factors contributing to COVID-19 severity.

Understanding sociodemographic factors behind COVID-19 severity relates to significant methodological difficulties, such as differences in testing policies and epidemics phase, as well as a large number of predictors that can potentially contribute to severity. To account for these difficulties, we assemble 115 predictors for more than 3,000 US counties and employ a well-defined COVID-19 severity measure derived from epidemiological dynamics modeling. We then use a number of advanced feature selection techniques from machine learning to determine which of these predictors significantly impact the disease severity. We obtain a surprisingly simple result, where only two variables are clearly and robustly selected-population density and proportion of African Americans. Possible causes behind this result are discussed. We argue that the approach may be useful whenever significant determinants of disease progression over diverse geographic regions should be selected from a large number of potentially important factors.

Eğitim İşleviyle Yenilenen Cibali Tütün ve Sigara Fabrikası’nda Çağdaş Mekan Kurguları

Space constitutes one of the important elements of the architectural discipline and gives meaning to a building by moving it beyond being a three-dimensional mass. Designing the spaces that the building contains and determining the actions they will contain makes architecture different from other building activities. The spaces in industrial buildings are designed in accordance with the use of large-scale machines where production will be provided. These structures, which remain in important positions in the city centers with the development of the cities, are abandoned by losing their functions with the changing production conditions. Due to their location, strong structural system, and high-quality structural features such as large volumes, giving these structures new functions and ensuring their survival creates an alternative building production technique. The structural features of industrial buildings enable various uses in function selection. The spaces needed in these buildings, which are renovated for reuse, are organized within the existing volumes with the help of space components and elements. In this study, additional designs made during the renovation of Cibali Tobacco and Cigarette Factory, which was used as a university education structure, to include contemporary education spaces were examined. Qualitative research method was used in the study, and the history, architectural features and renewed space design of Cibali Tobacco and Cigarette Factory, which constitutes the field study, were analyzed through written sources, photographs, and drawings. In the study, contemporary usage ideas in new spaces designed with the help of additional space components and elements were evaluated. As a result of the study, suggestions were made for additional designs to be made in industrial buildings to be re-functionalized.

Near Real-Time Social Distance Estimation In London

Abstract To mitigate the current COVID-19 pandemic, policy makers at the Greater London Authority, the regional governance body of London, UK, are reliant upon prompt, accurate and actionable estimations of lockdown and social distancing policy adherence. Transport for London, the local transportation department, reports they implemented over 700 interventions such as greater signage and expansion of pedestrian zoning at the height of the pandemic’s first wave with our platform providing key data for those decisions. Large well-defined heterogeneous compositions of pedestrian footfall and physical proximity are difficult to acquire, yet necessary to monitor city-wide activity (busyness) and consequently discern actionable policy decisions. To meet this challenge, we leverage our existing large-scale data processing urban air quality machine learning infrastructure to process over 900 camera feeds in near real-time to generate new estimates of social distancing adherence, group detection and camera stability. In this work, we describe our development and deployment of a computer vision and machine learning pipeline. It provides near immediate sampling and contextualization of activity and physical distancing on the streets of London via live traffic camera feeds. We introduce a platform for inspecting, calibrating and improving upon existing methods, describe the active deployment on real-time feeds and provide analysis over an 18 month period.

A Case for Micromachines in Laboratory-Based DFIG Wind Turbine Systems for Fault Studies

For laboratory-based systems, the selection of components is critical to ensure proper representation of the actual system. In past investigations, researchers made use of small-scale off-the-shelf wound rotor induction machines (WRIMs) to perform laboratory-based studies. However, there are no reports validating their applicability in comparison to using the actual machines. In this regard, this paper investigates the correlation between the dimensional parameters and the electrical asymmetry indicators for a large-scale wind generator and laboratory-based generators. For the latter, a comparison is made between using a standard small-scale off-the-shelf WRIM and a micromachine. It is shown that the harmonic frequencies in a healthy machine and those due to inherent asymmetries appear at the same positions on the stator current spectrum for the three machines. However, there is a difference in the amplitudes of these components, given their dependence on the winding factors for a specific machine. The dimensional analysis techniques used to scale down the large-scale generator into a micromachine ensure that these parameters are related. However, this is not the case for the standard small-scale WRIM. Further work demonstrates the correlation of the fault-related harmonics between the micromachine and the large-scale machine for a rotor inter-turn fault condition. Micromachines are designed to ensure that the time constants are proportional representations of the larger machine, thus providing an accurate reflection of what is expected under transient operating conditions of wind turbines. This would assist in the development of condition monitoring strategies that are suited for the actual systems in the industry.

Factors associated with resistance to SARS-CoV-2 infection discovered using large-scale medical record data and machine learning.

There have been over 621 million cases of COVID-19 worldwide with over 6.5 million deaths. Despite the high secondary attack rate of COVID-19 in shared households, some exposed individuals do not contract the virus. In addition, little is known about whether the occurrence of COVID-19 resistance differs among people by health characteristics as stored in the electronic health records (EHR). In this retrospective analysis, we develop a statistical model to predict COVID-19 resistance in 8,536 individuals with prior COVID-19 exposure using demographics, diagnostic codes, outpatient medication orders, and count of Elixhauser comorbidities in EHR data from the COVID-19 Precision Medicine Platform Registry. Cluster analyses identified 5 patterns of diagnostic codes that distinguished resistant from non-resistant patients in our study population. In addition, our models showed modest performance in predicting COVID-19 resistance (best performing model AUROC = 0.61). Monte Carlo simulations conducted indicated that the AUROC results are statistically significant (p < 0.001) for the testing set. We hope to validate the features found to be associated with resistance/non-resistance through more advanced association studies.

Programming for the near future: Concepts and pragmatic considerations

This article deals with the concept, architecture, and scientific-organizational problems of creating a new generation of integrated software intended for predictive modeling in engineering, energy, materials science, biology, medicine, economics, nature management, ecology, sociology, etc. Mathematical formulations include interdisciplinary direct and inverse extremely resource-intensive tasks, which are solved using computational methods and technologies of scalable parallelization by hybrid programming on heterogeneous supercomputers with distributed and hierarchical shared memory. The project concept includes the development of an instrumental computational environment that supports all stages of a large-scale machine experiment: geometric and functional modeling, generating of adaptive unstructured grids of various types and orders, approximation of initial equations, solution of emerging algebraic problems, postprocessing of the obtained results, optimization methods for inverse tasks, and machine learning and decision-making on the results of calculations. The effective functionality of the instrumented computing environment is based on high-performance computing and intelligent big data tools. The architecture of the instrumental computational environment provides for automated expansion of the composition of implemented models and applied algorithms, adaptation to the evolution of supercomputer platforms, user-friendly interfaces and active reuse of external software products, and coordinated participation of different groups of developers, which together should provide a long life cycle and demand for the created ecosystem by a wide range of users from different professional fields.

Federated Learning to Safeguard Patients Data: A Medical Image Retrieval Case

Healthcare data are distributed and confidential, making it difficult to use centralized automatic diagnostic techniques. For example, different hospitals hold the electronic health records (EHRs) of different patient populations; however, transferring this data between hospitals is difficult due to the sensitive nature of the information. This presents a significant obstacle to the development of efficient and generalizable analytical methods that require a large amount of diverse Big Data. Federated learning allows multiple institutions to work together to develop a machine learning algorithm without sharing their data. We conducted a systematic study to analyze the current state of FL in the healthcare industry and explore both the limitations of this technology and its potential. Organizations share the parameters of their models with each other. This allows them to reap the benefits of a model developed with a richer data set while protecting the confidentiality of their data. Standard methods for large-scale machine learning, distributed optimization, and privacy-friendly data analytics need to be fundamentally rethought to address the new problems posed by training on diverse networks that may contain large amounts of data. In this article, we discuss the particular qualities and difficulties of federated learning, provide a comprehensive overview of current approaches, and outline several directions for future work that are relevant to a variety of research communities. These issues are important to many different research communities.