AUTOMATIC ANALYSIS OF MICROSCOPY IMAGES USING THE DLGRAM01 CLOUD SERVICE

Исследование материалов методами микроскопии нередко включает стадию подсчета количества наблюдаемых объектов и определения их статистических параметров, для чего необходимо измерять сотни объектов. В работе описан облачный сервис DLgram01, который позволяет специалистам в области материаловедения, не имеющих навыков программирования, выполнять автоматизированную обработку изображений - определять количество и параметры (площадь, размер) изучаемых объектов. Сервис разработан с использованием новейших достижений в области глубокого машинного обучения, для обучения нейронной сети пользователю необходимо разметить несколько изучаемых объектов. Обучение нейронной сети производится автоматически за несколько минут. Важными особенностями сервиса DLgram01 является возможность корректировать результаты предсказания нейронной сети, а также получение детальной информации о всех распознанных объектах. Использование сервиса позволяет существенно сократить временные затраты на количественный анализ изображений, снизить влияние субъективного фактора, повысить точность анализа и его эргоемкость. The study of materials by microscopy often includes counting the number of observed objects and determining their statistical parameters, for which it is necessary to measure hundreds of objects. The created DLgram01 cloud service allows specialists in the field of materials science who do not have programming skills to perform automated image processing - to determine the number and parameters (area, size) of the objects under study. The service is developed using the latest achievements in the field of deep machine learning. To train a neural network, the user needs to label only several objects. The neural network is trained automatically in a few minutes. Important features of the DLgram01 service are the ability to adjust the results of neural network prediction, as well as obtaining detailed information about all recognized objects. Using the service allows to significantly decrease the time for quantitative image analysis, reduce the influence of the subjective factor, increase the accuracy of the analysis and its ergo-intensity.

cuSCNN: A Secure and Batch-Processing Framework for Privacy-Preserving Convolutional Neural Network Prediction on GPU

The emerging topic of privacy-preserving deep learning as a service has attracted increasing attention in recent years, which focuses on building an efficient and practical neural network prediction framework to secure client and model-holder data privately on the cloud. In such a task, the time cost of performing the secure linear layers is expensive, where matrix multiplication is the atomic operation. Most existing mix-based solutions heavily emphasized employing BGV-based homomorphic encryption schemes to secure the linear layer on the CPU platform. However, they suffer an efficiency and energy loss when dealing with a larger-scale dataset, due to the complicated encoded methods and intractable ciphertext operations. To address it, we propose cuSCNN, a secure and efficient framework to perform the privacy prediction task of a convolutional neural network (CNN), which can flexibly perform on the GPU platform. Its main idea is 2-fold: (1) To avoid the trivia and complicated homomorphic matrix computations brought by BGV-based solutions, it adopts GSW-based homomorphic matrix encryption to efficiently enable the linear layers of CNN, which is a naive method to secure matrix computation operations. (2) To improve the computation efficiency on GPU, a hybrid optimization approach based on CUDA (Compute Unified Device Architecture) has been proposed to improve the parallelism level and memory access speed when performing the matrix multiplication on GPU. Extensive experiments are conducted on industrial datasets and have shown the superior performance of the proposed cuSCNN framework in terms of runtime and power consumption compared to the other frameworks.

Investigation on the Displacement Ductility Coefficient of Reinforced Concrete Columns Strengthened with Textile-Reinforced Concrete

To evaluate the seismic performance of reinforced concrete (RC) columns strengthened with textile-reinforced concrete (TRC), based on the ABAQUS numerical analysis results of 15 TRC-strengthened RC columns, the grey correlation theory was used to determine the input variables of the model, and the accuracy of the numerical simulation results is verified by some experiments. Then, according to FEM data, a neural network prediction model was established for the displacement ductility coefficients of TRC-strengthened columns, and a formula was proposed for calculating the displacement ductility coefficient. The results showed that the BP (backpropagation) neural network model had good rationality and accuracy and that the ductility coefficients of the strengthened columns calculated by the model agreed well with the experimental values. Therefore, the model can be applied for predicting the displacement ductility coefficients of TRC-strengthened columns and can be used as a reference for engineering design.

Prediction of Short-Term Stock Price Trend Based on Multiview RBF Neural Network

Stock price prediction is important in both financial and commercial domains, and using neural networks to forecast stock prices has been a topic of ongoing research and development. Traditional prediction models are often based on a single type of data and do not account for the interplay of many variables. This study covers a radial basis neural network modeling technique with multiview collaborative learning capabilities for incorporating the impacts of numerous elements into the prediction model. This research offers a multiview RBF neural network prediction model based on the classic RBF network by integrating a collaborative learning item with multiview learning capabilities (MV-RBF). MV-RBF can make full use of both the internal information provided by the correlation between each view and the distinct characteristics of each view to form independent sample information. By using two separate stock qualities as input feature information for trials, this study proves the viability of the multiview RBF neural network prediction model on a real data set.

Application of BP Neural Network Based on Petrophysical Big Data Mining

Based on the big data mining method of petrophysical data, this paper studies the method and application of BP neural network to establish nonlinear interpretation model in distributed cloud computing environment. The nonlinear mapping relationship between the relative objective logging response and actual formation component is established by extracting the data mining result model, which overcomes existing deficiencies of the conventional logging interpretation procedure based on the homogeneity theory, linear hypothesis and the use of statistical experience simplifying model and parameters. The results show that network prediction model has been improved and has superior reference value for solving practical problems of interpretation under complex geological conditions.