Search

A separation of the flow region from the surroundings is an essential step in the analysis of smoke visualization images. The separation can be performed using several detection methods from the image segmentation group. This paper deals with the border detection of the air flow downstream of a benchmark automotive vent using different threshold-based detection methods. An assessment of the methods on the basis of the resulting image quality is also addressed. The quality level depends on the quantity and brightness of disturbances in the background area. The disturbance is usually an isolated region of smoke, which naturally cannot be a part of the flow. Three representative images of different quality levels were selected for the detection, and three methods were used for the evaluation. Each of the methods was used to determine the threshold differently (by the level, by the ratio, and by the change of brightness). It is demonstrated that the change-based method with an appropriately selected parameter is the most convenient for images with the worst quality level while level- and ratio-based methods are only applicable for images of good quality. A separation of the flow region from the surroundings is an essential step in the analysis of smoke visualization images. The separation can be performed using several detection methods from the image segmentation group. This paper deals with the border detection of the air flow downstream of a benchmark automotive vent using different threshold-based detection methods. An assessment of the methods on the basis of the resulting image quality is also addressed. The quality level depends on the quantity and brightness of disturbances in the background area. The disturbance is usually an isolated region of smoke, which naturally cannot be a part of the flow. Three representative images of different quality levels were selected for the detection, and three methods were used for the evaluation. Each of the methods was used to determine the threshold differently (by the level, by the ratio, and by the change of brightness). It is demonstrated that the change-based method with an appropriately selected parameter is the most convenient for images with the worst quality level while level- and ratio-based methods are only applicable for images of good quality.

The physics reach of the HL-LHC will be limited by how efficiently the experiments can use the available computing resources, i.e. affordable software and computing are essential. The development of novel methods for charged particle reconstruction at the HL-LHC incorporating machine learning techniques or based entirely on machine learning is a vibrant area of research. In the past two years, algorithms for track pattern recognition based on graph neural networks (GNNs) have emerged as a particularly promising approach. Previous work mainly aimed at establishing proof of principle. In the present document we describe new algorithms that can handle complex realistic detectors. The new algorithms are implemented in ACTS, a common framework for tracking software. This work aims at implementing a realistic GNN-based algorithm that can be deployed in an HL-LHC experiment., Comment: Paper submitted to vCHEP2021 (25th International Conference on Computing in High-Energy Physics and Nuclear Physics), 10 pages, 4 figures

Machine learning algorithms are gaining ground in high energy physics for applications in particle and event identification, physics analysis, detector reconstruction, simulation and trigger. Currently, most data-analysis tasks at LHC experiments benefit from the use of machine learning. Incorporating these computational tools in the experimental framework presents new challenges. This paper reports on the implementation of the end-to-end deep learning with the CMS software framework and the scaling of the end-to-end deep learning with multiple GPUs. The end-to-end deep learning technique combines deep learning algorithms and low-level detector representation for particle and event identification. We demonstrate the end-to-end implementation on a top quark benchmark and perform studies with various hardware architectures including single and multiple GPUs and Google TPU.

The usefulness and value of Multi-step Machine Learning (ML), where a task is organized into connected sub-tasks with known intermediate inference goals, as opposed to a single large model learned end-to-end without intermediate sub-tasks, is presented. Pre-optimized ML models are connected and better performance is obtained by re-optimizing the connected one. The selection of an ML model from several small ML model candidates for each sub-task has been performed by using the idea based on Neural Architecture Search (NAS). In this paper, Differentiable Architecture Search (DARTS) and Single Path One-Shot NAS (SPOS-NAS) are tested, where the construction of loss functions is improved to keep all ML models smoothly learning. Using DARTS and SPOS-NAS as an optimization and selection as well as the connections for multi-step machine learning systems, we find that (1) such a system can quickly and successfully select highly performant model combinations, and (2) the selected models are consistent with baseline algorithms, such as grid search, and their outputs are well controlled.

With the increasing needs of accurate simulation, the 3-D diffusion reactor physics module has been implemented in HTGR’s engineering simulator to give better neutron dynamics results instead of point kinetics model used in previous nuclear power plant simulators. As the requirement of real-time calculation of nuclear power plant simulator, the cross-sections used in 3-D diffusion module must be calculated very efficiently. Normally, each cross-section in simulator is calculated in the form of polynomial by function of several concerned variables, the expression of which was finalized by multivariate regression from large number scattered database generated by previous calculation. Since the polynomial is explicit and prepared in advance, the cross-sections could be calculated quickly enough in running simulator and achieve acceptable accuracy especially in LWR simulations. However, some of concerned variables in HTGR are in large scope and also the relationships of these variables are non-linear and very complex, it is very hard to use polynomial to meet full range accuracy. In this paper, a cross-section generating method used in HTGR simulator is proposed, which is based on machine learning methods, especially deep neuron network and tree regression methods. This method first uses deep neuron networks to consider the nonlinear relationships between different variables and then uses a tree regression to achieve accurate cross-section results in full range, the parameters of deep neuron networks and tree regression are learned automatically from the scattered database generated by VSOP. With the numerical tests, the proposed cross-section generating method could get more accurate cross-section results and the calculation time is acceptable by the simulator.

Computer paralinguistic analysis is widely used in security systems, biometric research, call centers and banks. Paralinguistic models estimate different physical properties of voice, such as pitch, intensity, formants and harmonics to classify emotions. The main goal is to find such features that would be robust to outliers and will retain variety of human voice properties at the same time. Moreover, the model used must be able to estimate features on a time scale for an effective analysis of voice variability. In this paper a paralinguistic model based on Bidirectional Long Short-Term Memory (BLSTM) neural network is described, which was trained for vocal-based emotion recognition. The main advantage of this network architecture is that each module of the network consists of several interconnected layers, providing the ability to recognize flexible long-term dependencies in data, which is important in context of vocal analysis. We explain the architecture of a bidirectional neural network model, its main advantages over regular neural networks and compare experimental results of BLSTM network with other models.

Crop losses are a major threat to the wellbeing of rural families, to the economy and governments, and to food security worldwide. The goal of our research is to develop a multi-functional platform to help the farming community to tilt against plant diseases. In our previous works, we reported about the creation of a special database of healthy and diseased plants’ leaves consisting of five sets of grapes images and proposed a special classification model based on a deep siamese network followed by k-nearest neighbors (KNN) classifier. Then we extended our database to five sets of images for grape, corn, and wheat – 611 images in total. Since after this extension the classification accuracy decreased to 86 %, we propose in this paper a novel architecture with a deep siamese network as feature extractor and a single-layer perceptron as a classifier that results in a significant gain of accuracy, up to 96 %.

One of the most important aspects of data processing at flavor physics experiments is the particle identification (PID) algorithm. In LHCb, several different sub-detector systems provide PID information: the Ring Imaging Cherenkov detectors, the hadronic and electromagnetic calorimeters, and the muon chambers. The charged PID based on the sub-detectors response is considered as a machine learning problem solved in different modes: one-vs-rest, one-vs-one and multi-classification, which affect the models training and prediction. To improve charged particle identification for pions, kaons, protons, muons and electrons, neural network and gradient boosting models have been tested. This paper presents these models and their performance evaluated on Run 2 data and simulation samples. A discussion of the performances is also presented.

We present a catalogue of galaxy photometric redshifts for the Sloan Digital Sky Survey (SDSS) Data Release 12. We use various supervised learning algorithms to calculate redshifts using photometric attributes on a spectroscopic training set. Two training sets are analysed in this paper. The first training set consists of 995,498 galaxies with redshifts up to $z \approx 0.8$. On the first training set, we achieve a cost function of 0.00501 and a root mean squared error value of 0.0707 using the XGBoost algorithm. We achieved an outlier rate of 2.1\% and 86.81\%, 95.83\%, 97.90\% of our data points lie within one, two, and three standard deviation of the mean respectively. The second training set consists of 163,140 galaxies with redshifts up to $z\approx0.2$ and is merged with the Galaxy Zoo 2 full catalog. We also experimented on convolutional neural networks to predict five morphological features (Smooth, Features/Disk, Star, Edge-on, Spiral). We achieve a root mean squared error of 0.117 when validated against an unseen dataset with over 200 epochs. Morphological features from the Galaxy Zoo, trained with photometric features are found to consistently improve the accuracy of photometric redshifts., Comment: Poster presented at the XLVIII International Symposium on Multiparticle Dynamics (ISMD2018)

The macro-scale behavior of granular materials is strongly influenced by grain kinematics. The mobility of the grains in turn is affected by grain morphology which needs to be comprehensively characterized. Initially, morphological descriptors were determined by manual processes which were tedious and cumbersome. But with the help of image processing techniques and computational geometry, this problem can be handled with ease. The long-established descriptors such as sphericity and roundness can be estimated by operating over the grain boundary obtained from 2D images of the particle. The main objective of this study is to quantify these descriptors in a computationally effective way. The roundness quantification in previous studies involves the sequence of processes such as removing the noise, corner identification and fitting circles. This paper details the necessary modifications to the quantification process required to reduce the cost of time for a single particle. Further, the influence of different smoothing techniques and a new corner identification method will be detailed.

The research conducted in this paper is in the field of machine learning. The main object of the research is the learning process of an artificial neural network in order to increase its efficiency. The algorithm based on the analysis of retrospective learning data. The dynamics of changes in the values of the weights of an artificial neural network during training is an important indicator of training efficiency. The algorithm proposed in this work is based on changing the weight gradients values. Changing of the gradients weights makes it possible to understand how actively the network weights change during training. This knowledge helps to diagnose the training process and makes an adjusting the training parameters. The results of the algorithm can be used to train an artificial neural network. The network will help to determine the set of measures (actions) needed to optimize the learning process by the algorithm results.

The Liquid Argon Time Projection Chamber (LArTPC) technology plays an essential role in many current and future neutrino experiments. Accurate and fast simulation is critical to developing efficient analysis algorithms and precise physics model projections. The speed of simulation becomes more important as Deep Learning algorithms are getting more widely used in LArTPC analysis and their training requires a large simulated dataset. Heterogeneous computing is an efficient way to delegate computing-heavy tasks to specialized hardware. However, as the landscape of the compute accelerators is evolving fast, it becomes more and more difficult to manually adapt the code constantly to the latest hardware or software environments. A solution which is portable to multiple hardware architectures while not substantially compromising performance would be very beneficial, especially for long-term projects such as the LArTPC simulations. In search of a portable, scalable and maintainable software solution for LArTPC simulations, we have started to explore high-level portable programming frameworks that support several hardware backends. In this paper, we will present our experience porting the LArTPC simulation code in the Wire-Cell toolkit to NVIDIA GPUs, first with the CUDA programming model and then with a portable library called Kokkos. Preliminary performance results on NVIDIA V100 GPUs and multi-core CPUs will be presented, followed by a discussion of the factors affecting the performance and plans for future improvements., 10 pages, vCHEP2021

High-energy physics computing is a typical data-intensive calculation. Each year, petabytes of data needs to be analyzed, and data access performance is increasingly demanding. The tiered storage system scheme for building a unified namespace has been widely adopted. Generally, data is stored on storage devices with different performances and different prices according to different access frequency. When the heat of the data changes, the data is then migrated to the appropriate storage tier. At present, heuristic algorithms based on artificial experience are widely used in data heat prediction. Due to the differences in computing models of different users, the accuracy of prediction is low. A method for predicting future access popularity based on file access characteristics with the help of LSTM deep learning algorithm is proposed as the basis for data migration in hierarchical storage. This paper uses the real data of high-energy physics experiment LHAASO as an example for comparative testing. The results show that under the same test conditions, the model has higher prediction accuracy and stronger applicability than existing prediction models.

This paper introduces a generic and scalable anomaly detection framework. Anomaly detection can improve operation and maintenance efficiency and assure experiments can be carried out effectively. The framework facilitates common tasks such as data sample building, retagging and visualization, deviation measurement and performance measurement for machine learning-based anomaly detection methods. The samples we used are sourced from Ganglia monitoring data. There are several anomaly detection methods to handle spatial and temporal anomalies within the framework. Finally, we show the rudimental application of the framework on Lustre distributed file systems in daily operation and maintenance.

Recently, by using deep learning methods, a computer is able to surpass or come close to matching human performance on image analysis and recognition. This advanced methods could also help extracting features from neutron scattering experimental data. Those data contain rich scientific information about structure and dynamics of materials under investigation. Deep learning could help researchers better understand the link between experimental data and materials properties. Moreover,it could also help to optimize neutron scattering experiment by predicting the best possible instrument configuration. Among all possible experimental methods, we begin our study on the small-angle neutron scattering (SANS) data and by predicting the structure geometry of the sample material at an early stage. This step is a keystone to predict the experimental parameters to properly setup the instrument as well as the best measurement strategy. In this paper, we propose to use transfer learning to retrain a convolutional neural networks (CNNs) based pre rained model to adapt the scattering images classification, which could predict the structure of the materials at an early stage in the SANS experiment. This deep neural network is trained and validated on simulated database, and tested on real scattering images.

AI and deep learning are experiencing explosive growth in almost every domain involving analysis of big data. Deep learning using Deep Neural Networks (DNNs) has shown great promise for such scientific data analysis applications. However, traditional CPU-based sequential computing without special instructions can no longer meet the requirements of mission-critical applications, which are compute-intensive and require low latency and high throughput. Heterogeneous computing (HGC), with CPUs integrated with GPUs, FPGAs, and other science-targeted accelerators, offers unique capabilities to accelerate DNNs. Collaborating researchers at SHREC1at the University of Florida, CERN Openlab, NERSC2at Lawrence Berkeley National Lab, Dell EMC, and Intel are studying the application of heterogeneous computing (HGC) to scientific problems using DNN models. This paper focuses on the use of FPGAs to accelerate the inferencing stage of the HGC workflow. We present case studies and results in inferencing state-of-the-art DNN models for scientific data analysis, using Intel distribution of OpenVINO, running on an Intel Programmable Acceleration Card (PAC) equipped with an Arria 10 GX FPGA. Using the Intel Deep Learning Acceleration (DLA) development suite to optimize existing FPGA primitives and develop new ones, we were able accelerate the scientific DNN models under study with a speedup from 2.46x to 9.59x for a single Arria 10 FPGA against a single core (single thread) of a server-class Skylake CPU.

Software defect prediction is an activity that aims at narrowing down the most likely defect-prone software modules and helping developers and testers to prioritize inspection and testing. This activity can be addressed by using Machine Learning techniques applied to software metrics datasets that are usually unlabelled, i.e. they lack modules classification in terms of defectiveness. To overcome this limitation, in addition to the usual data pre-processing operations to manage mission values and/or to remove inconsistencies, researches have to adopt an approach to label their unlabelled software datasets. The extraction of defectiveness data to label all the instances of the datasets is an extremely time and effort consuming operation. In literature, many studies have introduced approaches to build a defect prediction models on unlabelled datasets.In this paper, we describe the analysis of new unlabelled datasets from WLCG software, coming from HEP-related experiments and middleware, by using Machine Learning techniques. We have experimented new approaches to label the various modules due to the heterogeneity of software metrics distribution. We discuss a number of lessons learned from conducting these activities, what has worked, what has not and how our research can be improved.

In this research, we investigated two approaches to detect job anomalies and/or contention for large scale computing efforts: 1. Preemptive job scheduling using binomial classification long short-term memory networks 2. Forecasting intra-node computing loads from the active jobs and additional job(s) For approach 1, we achieved a 14% improvement in computational resources utilization and an overall classification accuracy of 85% on real tasks executed in a High Energy Physics computing workflow. For this paper, we present the preliminary results used in second approach.

This paper proposes a method for reconstructing a depth map obtained using a stereo pair image. The proposed approach is based on a geometric model for the synthesis of patches. The entire image is preliminarily divided into blocks of different size, where large blocks are used to restore homogeneous areas, and small blocks are used to restore details of the image structure. Lost pixels are recovered by copying the pixel values from the source based on the similarity criterion. We used a trained neural network to select the “best like” patch. Experimental results show that the proposed method gives better results than other modern methods, both in subjective and objective measurements for reconstructing a depth map.

Magnetorheological Finishing (MRF) is widely regarded as an effective technique to polish and figure aspheric optics. MRF based on Robot Arm is developed by us. This new machine is more flexible, efficient, cost-effective and smaller space-usage in optical shop, compared with the traditional MRF machines. The components of MRF based on Robot Arm are introduced firstly. Position-attitude control and polishing tool path are also studied in this paper. The experiments and application of MRF based on Robot Arm demonstrate the effectiveness and validity of MRF based on Robot Arm in optical fabrication.

The problem of optimally activating the flexible energy sources (short- and long-term storage capacities) of electricity microgrid is formulated as a sequential decision making problem under uncertainty where, at every time-step, the uncertainty comes from the lack of knowledge about future electricity consumption and weather dependent PV production. This paper proposes to address this problem using deep reinforcement learning. To this purpose, a specific deep learning architecture has been used in order to extract knowledge from past consumption and production time series as well as any available forecasts. The approach is empirically illustrated in the case of off-grid microgrids located in Belgium and Russia.

In this paper, we investigate the possibility of applying various approaches to solving the problem of medium-term forecasting of daily passenger traffic volumes in the Moscow metro (MM): 1) on the basis of artificial neural networks (ANN); 2) using the singular-spectral analysis implemented in the package “Caterpillar”-SSA; 3) sharing the ANN and the “Caterpillar”-SSA approach. We demonstrate that the developed methods and algorithms allow us to conduct medium-term forecasting of passenger traffic in the MM with reasonable accuracy.

This paper will focus on the study and identifying various threshold values for two commonly used edge detection techniques, which are Sobel and Canny Edge detection. The idea is to determine which values are apt in giving accurate results in identifying a particular leukemic cell. In addition, evaluating suitability of edge detectors are also essential as feature extraction of the cell depends greatly on image segmentation (edge detection). Firstly, an image of M7 subtype of Acute Myelocytic Leukemia (AML) is chosen due to its diagnosing which were found lacking. Next, for an enhancement in image quality, noise filters are applied. Hence, by comparing images with no filter, median and average filter, useful information can be acquired. Each threshold value is fixed with value 0, 0.25 and 0.5. From the investigation found, without any filter, Canny with a threshold value of 0.5 yields the best result.

An important element of UAV equipment are observation heads. Observation head is multisensor optoelectronic system. It includes TV camera and infrared camera, safe for the eye rangefinder and laser pointer and a navigation system GPS. Image stability is provided by electro-mechanical gyro stability system. Optoelectronic head are designed for use in systems of detection, identification and observation of ground and air targets and in daylight and at night. In this paper simulation studies of the air masses flow around the observation head placed on the UAV deck, for different speeds of UAV movement were conducted. The results of numerical research were presented in a graphical form.

In the paper it is shown that at present the problem of measurements of emotions as multidimensional (multiparametric) quantities is of interest. A measurement model developed by the authors is presented. It enables signals-stimuli generating emotions in sound combinations and their series to be selected and quantitatively estimated. The measurement model provides an adjusted delay (short-time memory) and joint non-linear conversion of audible and delayed acoustical signals, selection of infrasound signals-stimuli as well as identification of emotional images including the ensembles of signals-stimuli and their series. It is demonstrated how the “mechanism” of emotional perception of sound impacts was developed in the process of evolution. Possibilities to decode the emotional content of biolinguistic signals of animals are considered. Future trends of investigations are indicated.

The Digital Image Correlation (DIC) technique enables full field, noncontact measurements of displacements and strains of a wide variety of objects. An adaptation of the DIC technique for monitoring of civil-engineering structures is presented in the paper. A general concept of the complex, automatic monitoring system, in which the DIC sensor plays an important role is described. Some new software features, which aim to facilitate outdoor measurements and speed up the correlation analysis, is also introduced. As an example of application, measurements of a railway bridge in Nieporet (Poland) are presented. The experimental results are compared with displacements of a FEM model of the bridge.

Holographic interferometry (HI) techniques are powerful tools for experimental mechanics which can be used for displacement fields assessment on a loaded object surface. Used with no contact and high resolution HI can be combined with image processing methods for data analysis. In these techniques the speckle should be fully resolved by the image detector to record an interferogram. This leads to noisy recordings which makes it's post-processing much difficult. Each measurement with HI ends normally in a phase map computation representing the object displacement. Phase calculation algorithms are based on spatial or temporal phase modulation and lead to wrapped phase maps. To obtain the continuous phase distribution several solutions have been implemented. This paper describes the main filtering and unwrapping algorithms available and compares it's performance to solve discontinuities in noisy recordings.