Wyniki wyszukiwania - BazTech

1

Automatic diagnosis of severity of COVID-19 patients using an ensemble of transfer learning models with convolutional neural networks in CT images

Shalbaf Ahmad, Gifani Parisa, Mehri-Kakavand Ghazal, Pursamimi Mohamad, Ghorbani Mahdi, Davanloo Amirhossein Abbaskhani, Vafaeezadeh Majid

Polish Journal of Medical Physics and Engineering

|

2022

|

Vol. 28, Iss. 3

117--126

EN

Introduction: Quantification of lung involvement in COVID-19 using chest Computed tomography (CT) scan can help physicians to evaluate the progression of the disease or treatment response. This paper presents an automatic deep transfer learning ensemble based on pre-trained convolutional neural networks (CNNs) to determine the severity of COVID -19 as normal, mild, moderate, and severe based on the images of the lungs CT. Material and methods: In this study, two different deep transfer learning strategies were used. In the first procedure, features were extracted from fifteen pre-trained CNNs architectures and then fed into a support vector machine (SVM) classifier. In the second procedure, the pre-trained CNNs were fine-tuned using the chest CT images, and then features were extracted for the purpose of classification by the softmax layer. Finally, an ensemble method was developed based on majority voting of the deep learning outputs to increase the performance of the recognition on each of the two strategies. A dataset of CT scans was collected and then labeled as normal (314), mild (262), moderate (72), and severe (35) for COVID-19 by the consensus of two highly qualified radiologists. Results: The ensemble of five deep transfer learning outputs named EfficientNetB3, EfficientNetB4, InceptionV3, NasNetMobile, and ResNext50 in the second strategy has better results than the first strategy and also the individual deep transfer learning models in diagnosing the severity of COVID-19 with 85% accuracy. Conclusions: Our proposed study is well suited for quantifying lung involvement of COVID-19 and can help physicians to monitor the progression of the disease.

2

Deep neural network and ANN ensemble for slope stability prediction

Gupta A., Aggarwal Y., Aggarwal P.

Archives of Materials Science and Engineering

|

2022

|

Vol. 116, nr 1

14--27

EN

Purpose: Application of deep neural networks (DNN) and ensemble of ANN with bagging for estimating of factor of safety (FOS) of soil stability with a comparative performance analysis done for all techniques. Design/methodology/approach: 1000 cases with different geotechnical and similar Geometrical properties were collected and analysed using the Limit Equilibrium based Morgenstern-Price Method with input variables as the strength parameters of the soil layers, i.e., Su (Upper Clay), Su (Lower Clay), Su (Peat), angle of internal friction (φ), Su (Embankment) with the factor of safety (FOS) as output. The evaluation and comparison of the performance of predicted models with cross-validation having ten folds were made based on correlation-coefficient (CC), Nash-Sutcliffe-model efficiency-coefficient (NSE), root-mean-square-error (RMSE), mean-absolute-error (MAE) and scattering-index (S.I.). Sensitivity analysis was conducted for the effects of input variables on FOS of soil stability based on their importance. Findings: The results showed that these techniques have great capability and reflect that the proposed model by DNN can enhance performance of the model, surpassing ensemble in prediction. The Sensitivity analysis outcome demonstrated that Su (Lower Clay) significantly affected the factor of safety (FOS), trailed by Su (Peat). Research limitations/implications: This paper sets sight on use of deep neural network (DNN) and ensemble of ANN with bagging for estimating of factor of safety (FOS) of soil stability. The current approach helps to understand the tangled relationship of various inputs to estimate the factor of safety of soil stability using DNN and ensemble of ANN with bagging. Practical implications: A dependable prediction tool is provided, which suggests that model can help scientists and engineers optimise FOS of soil stability. Originality/value: Recently, DNN and ensemble of ANN with bagging have been used in various civil engineering problems as reported by several studies and has also been observed to be outperforming the current prevalent modelling techniques. DNN can signify extremely changing and intricate high-dimensional functions in correlation to conventional neural networks. But on a detailed literature review, the application of these techniques to estimate factor of safety of soil stability has not been observed.

3

A Three Dimensional Empirical Study of Logging Questions From Six Popular Q & A Websites

Gujral Harshit, Sharma Abhinav, Lal Sangeeta, Kumar Lov

e-Informatica Software Engineering Journal

|

2019

|

Vol. 13, nr 1

105--139

EN

Background: Q&A websites such as StackOverflow or Serverfault provide an open platform for users to ask questions and to get help from experts present worldwide. These websites not only help users by answering their questions but also act as a knowledge base. These data present on these websites can be mined to extract valuable information that can benefit the software practitioners. Software engineering research community has already understood the potential benefits of mining data from Q&A websites and several research studies have already been conducted in this area. Aim: The aim of the study presented in this paper is to perform an empirical analysis of logging questions from six popular Q&A websites. Method: We perform statistical, programming language and content analysis of logging questions. Our analysis helped us to gain insight about the logging discussion happening in six different domains of the StackExchange websites. Results: Our analysis provides insight about the logging issues of software practitioners: logging questions are pervasive in all the Q&A websites, the mean time to get accepted answer for logging questions on SU and SF websites are much higher as compared to other websites, a large number of logging question invite a great amount of discussion in the SoftwareEngineering Q&A website, most of the logging issues occur in C++ and Java, the trend for number of logging questions is increasing for Java, Python, and Javascript, whereas, it is decreasing or constant for C, C++, C#, for the ServerFault and Superuser website 'C' is the dominant programming language.

4

Beauty of historic urban centres - evolution in conservation theory

Bardzińska-Bonenberg Teresa, Bączkowska Magdalena

Architecture Civil Engineering Environment

|

2019

|

Vol. 12, no. 1

7--17

EN

The paper aims to investigate if the changing principles in the protection of historic urban spaces arose as a result of evolving rationale of modern architecture and town planning over the last two centuries. The research was performed using a chronological comparison of the literature, source texts and analyses of graphic materials. Considerations include the concept of beauty of a city upgraded by its reconstruction and conservation, in the perspective of the 19th century great theoreticians representing two different aesthetic attitudes: Viollet-le-Duc and John Ruskin. The theory concerning beauty as an essential value of a city underwent a radical change in the 2nd half of the 20th century. The initial theories were followed by expert groups and institutions resulting in formulating directives and charters, e.g. the Venice Charter. Conventions and regulations concerning the protection of architectural and urban heritage were the achievement of communities, also beyond Europe. They expanded the concept of beauty of a city by the aspect of cultural heritage, taking into account the human being and the role of aesthetic experience (Cesare Brandi). The research reveals a new understanding of historic urban centres. Starting with monuments protection, now conservation encompasses heritage spaces as a whole, implements legal provisions and often influences development of new methods and technologies with social aspect: sustainability and preservation of cultural continuity.

5

A Novel Framework for Reliable Network Prediction of Small Scale Wireless Sensor Networks (SSWSNs)

Sandhu J. K., Verma A. K., Rana P. S.

Fundamenta Informaticae

|

2018

|

Vol. 160, nr 3

303--341

EN

In Small Scale Wireless Sensor Networks (SSWSNs), reliability is defined as the capability of a network to perform its intended task under certain conditions for a stated time span. There are many tools for modeling and analyzing the reliability of a network. As the intricacy of various networks is increasing, there is a need for many sophisticated methods for reliability analysis. The term reliability is used as an umbrella term to capture various attributes such as safety, availability, security, and ease of use. The existing methods have many shortcomings which include inadequacy of a novel framework and inefficacy to handle scalable networks. This paper presents a novel framework which predicts the overall reliability of the SSWSNs in terms of performance metrics such as, sent packets, received packets, packets forfeit, packet delivery ratio and throughput. This framework includes various phases starting with scenario generation, construction of a dataset, applying ensemble based machine learning techniques to predict the parameters which cannot be calculated. The ensemble model predicts with an optimum accuracy of 99.9% for data flow, 99.9% for the protocol used and 97.6% for the number of nodes. Finally, to check the robustness of the ensemble model 10-fold cross-validation is used. The dataset used in this work is available as a supplement at http://bit.ly/SSWSN-Reliability.

6

A hybrid gene selection method for microarray recognition

Shukla A. K., Singh P., Vardhan M.

Biocybernetics and Biomedical Engineering

|

2018

|

Vol. 38, no. 4

975--991

EN

DNA microarray data is expected to be a great help in the development of efficient diagnosis and tumor classification. However, due to the small number of instances compared to a large number of genes, many of the computational learning methods encounter difficulties to select the low subgroups. In order to select significant genes from the high dimensional data for tumor classification, nowadays, several researchers are exploring microarray data using various gene selection methods. However, there is no agreement between existing gene selection techniques that produce the relevant gene subsets by which it improves the classification accuracy. This motivates us to invent a new hybrid gene selection method which helps to eliminate the misleading genes and classify a disease correctly in less computational time. The proposed method composes of two-stage, in the first stage, EGS method using multi-layer approach and f-score approach is applied to filter the noisy and redundant genes from the dataset. In the second stage, adaptive genetic algorithm (AGA) work as a wrapper to identify significant genes subsets from the reduced datasets produced by EGS that can contribute to detect cancer or tumor. AGA algorithm uses the support vector machine (SVM) and Naïve Bayes (NB) classifier as a fitness function to select the highly discriminating genes and to maximize the classification accuracy. The experimental results show that the proposed framework provides additional support to a significant reduction of cardinality and outperforms the state-of-art gene selection methods regarding accuracy and an optimal number of genes.

7

Classic palace architecture of the eastern Podillia in Ukraine: historical background and key factors

Khorosha O., Smoliak V.

Architecture Civil Engineering Environment

|

2018

|

Vol. 11, no. 3

13--25

EN

The article deals with the palace architecture of the style of classicism in the Eastern Podillia in Ukraine of the late 18th and early 20th centuries, in the context of historical and socio-political backgrounds. This issue remains poorly studied, because it was not considered comprehensively for Eastern Podillia. The purpose of the article is to establish the factors and historical preconditions of the architecture of the palace complexes in accordance with the world tendencies and regional features of Eastern Podillia. The complex of general scientific and special research methods was applied in the work. It was established, that the palace architecture of Eastern Podillia was formed dependent on socio-economic and political changes in society, a complex of the main factors and worldview-cultural desires of the owners of estates. Determined, that the principles of the European classicism had an impact on the creation of conceptual features and typical signs of the palace complexes of the Eastern Podilia in the context of regional architectural tendencies and belonging to the territory of the PolishLithuanian Commonwealth in the 18 century.

8

Characterization of symbolic rules embedded in deep DIMLP networks : a challenge to transparency of deep learning

Bologna G., Hayashi Y.

Journal of Artificial Intelligence and Soft Computing Research

|

2017

|

Vol. 7, No. 4

265--286

EN

Rule extraction from neural networks is a fervent research topic. In the last 20 years many authors presented a number of techniques showing how to extract symbolic rules from Multi Layer Perceptrons (MLPs). Nevertheless, very few were related to ensembles of neural networks and even less for networks trained by deep learning. On several datasets we performed rule extraction from ensembles of Discretized Interpretable Multi Layer Perceptrons (DIMLP), and DIMLPs trained by deep learning. The results obtained on the Thyroid dataset and the Wisconsin Breast Cancer dataset show that the predictive accuracy of the extracted rules compare very favorably with respect to state of the art results. Finally, in the last classification problem on digit recognition, generated rules from the MNIST dataset can be viewed as discriminatory features in particular digit areas. Qualitatively, with respect to rule complexity in terms of number of generated rules and number of antecedents per rule, deep DIMLPs and DIMLPs trained by arcing give similar results on a binary classification problem involving digits 5 and 8. On the whole MNIST problem we showed that it is possible to determine the feature detectors created by neural networks and also that the complexity of the extracted rulesets can be well balanced between accuracy and interpretability.

9

On Seeking Consensus Between Document Similarity Measures

Kłopotek M.

Fundamenta Informaticae

|

2017

|

Vol. 156, nr 1

43--68

EN

This paper investigates the application of consensus clustering and meta-clustering to the set of all possible partitions of a data set. We show that when using a ”complement” of Rand Index as a measure of cluster similarity, the total-separation partition, putting each element in a separate set, is chosen.

10

Baza danych dla zespołowych atomowych skal czasu TA(PL) i UTC(PL)

Marszalec M., Nerkowski D.

Przegląd Telekomunikacyjny + Wiadomości Telekomunikacyjne

|

2009

|

nr 10

1960-1964

PL

Opisano bazę danych dla atomowych skal czasu TA(PL) oraz UTC(PL). Przedstawiono genezę powstania systemu porównań w Polsce oraz niezależnej Polskiej Atomowej Skali Czasu TA(PL). Opisano organizację systemu porównań atomowych wzorców czasu. Podano podstawy teoretyczne dotyczące zespołowych (grupowych) skal czasu oraz narzędzi charakteryzujących opis stanu wzorców. W dalszej części przedstawiono ogólny wygląd bazy wraz z wynikami działania zaimplementowanych w niej algorytmów. Na zakończenie przedstawiono kierunki dalszego rozwoju prac nad bazą.

EN

This paper presents Database for TA (PL) and UTC (PL). At the beginning it is described the genesis of the creation of a time standards comparison system in Poland and start of the independent Polish Atomic Time Scale TA (PL). Then it is shown the actual state of comparison system. Next part presents the theoretical basis for ensembles (team/group time scales) and tools to characterize the state of time standards. Following section gives a general description of Database, together with the results of implemented algorithms. At the end it is described the potential future development of Database.

11

Three-dimensional simulation of cavitation flow effects in pipleline system of turbomachines

Likhachev D, Kulagin V

Zeszyty Naukowe. Cieplne Maszyny Przepływowe - Turbomachinery / Politechnika Łódzka

|

2008

|

nr 133

213--220

EN

This paper presents a three-dimensional simulation of cavitation bubbles cloud effects on accessories and elements of compressor refrigerating installation, intermediate heat exchanger, as well as pumps and pumping stations. Bubbles appear on sections of these units due to fluid movement. Simulation has demonstrated high dependency of cavitation effects on a cavity shape. Thus, structural solutions subjected to change shapes of cavity can allow minimal cavitation effects on accessories, and elements of hydraulic equipment.

12

Learning from Skewed Class Multi-relational Databases

Guo H., Viktor H.L.

Fundamenta Informaticae

|

2008

|

Vol. 89, nr 1

69-94

EN

Relational databases, with vast amounts of data–from financial transactions, marketing surveys, medical records, to health informatics observations– and complex schemas, are ubiquitous in our society. Multirelational classification algorithms have been proposed to learn from such relational repositories, where multiple interconnected tables (relations) are involved. These methods search for relevant features both from a target relation (in which each tuple is associated with a class label) and relations related to the target, in order to better classify target relation tuples. However, in many practical database applications, such as credit card fraud detection and disease diagnosis, the target tuples are highly imbalanced. That is, the number of examples of one class (majority class) in the target relation is much higher than the others (minority classes). Many existing methods thus tend to produce poor predictive performance over the underrepresented class in the data. This paper presents a strategy to deal with such imbalanced multirelational data. The method learns from multiple views (feature sets) of relational data in order to construct view learners with different awareness of the imbalanced problem. These different observations possessed by multiple view learners are then combined, in order to yield a model which has better knowledge on both the majority and minority classes in a relational database. Experiments performed on six benchmarking data sets show that the proposed method achieves promising results when compared with other popular relational data mining algorithms, in terms of the ROC curve and AUC value obtained. In particular, an important result indicates that the method is superior when the class imbalanced is very high.