IT Research center

1. Fill in the ordering form

All customers must fill in an ordering form* and read general conditions for using IT Research center services.

• General conditions for using IT research center services
• Ordering form
• Users' list (must be included if there will be more than 3 users per one ordering form)

* Students must provide a form signed by his/her lecturer.

2. Sign and deliver

• Send a scanned document to: itapc@mif.vu.lt
• If jou are an external user contact itapc@mif.vu.lt

3. Acknowledgment

We would be very grateful if the scientists and researchers, who have used our supercomputer resources for research or experimental purposes, would mention IT Research center in the acknowledgement of the articles that will be published in scientific and popular science journals. For your convenience we offer a couple of examples:

• “The authors are thankful for the high performance computing resources provided by the Information Technology Research Center of Vilnius University.”
• “The authors are thankful for the HPC resources provided by the IT Research Center of Vilnius University.”

This study explores the feasibility of employing deep-learning models to perform approximate live judging for 4-way formation skydiving, a discipline where skydivers perform predefined formations in free fall. Given the time and effort required to judge these formations, our research aims to explore a more instantaneous method for approximate scoring. We created a novel dataset from twelve skydiving competitions and the first six skydiving formations, resulting in 2,946 frames annotated with formation names. Read article.

Institute of Data Science and Digital Technologies
PhD Olga Kurasova
PhD student Algimantas Skuodis

The separation or suppression of background noise in a given recording to extract clear speech benefits humans with hearing aids and devices such as Automatic Speech Recognition systems. In recent years, various experiments have been conducted to develop a Speech Enhancement (SE) model or framework that would produce high-quality estimations based on classical algorithms or Deep Neural Network techniques. This work investigates the most promising SE approaches, by replicating the experiments and using accurate training and testing criteria and the same set of evaluation metrics, to determine what produces better estimates of the clean speech signal. The authors are thankful for the high-performance computing resources provided by the Information Technology Open Access Center at the Faculty of Mathematics and Informatics of Vilnius University Information Technology Research Center. Read article.

Institute of Data Science and Digital Technologies
Dr. Povilas Treigys,
Student postgraduate Dominykas Stankevičius

Early detection of COVID-19 is of vital importance. The group of scientists’ study is aimed to automatically detect COVID-19 disease from CXR (chest radiography imaging) images while maximizing the accuracy in detection. Their proposed framework involves image enhancement technique using dataset which consists of three classes: COVID-19, viral pneumonia and normal CXR imagery. To establish the robustness of the proposed model, stratified 5-fold cross-validation is carried out. The results suggest that image enhancement, segmentation and future fusion methods used in this study could indeed improve classification results. The scientists are thankful for the high-performance computing resources provided by the Information Technology Open Access Center at the Faculty of Mathematics and Informatics of Vilnius University Information Technology Research Center. Read article.

The scientists of Vilnius University

Rising numbers and sophistication of security threats in the digital domain cause an increase in the demand for skilled cybersecurity professionals. In response, cybersecurity exercises, and in particular—cyber defence exercises (CDX) are becoming ever more popular. The goal of our research is to provide a proper methodology to optimise the exercises so that every team and each participant, including a non-technical trainee, are adequately evaluated and trained using the allocated resources most effectively. The authors are thankful for the high-performance computing resources provided by the Information Technology Open Access Center at the Faculty of Mathematics and Informatics of Vilnius University Information Technology Research Center. Read article.

The scientists of Cyber Security Lab

The group of scientists researched forecasting methods and modeling techniques which are important tools for the situations caused by COVID-19 virus. The main purpose of this study is to obtain short-term forecasts of disease epidemiology that could be useful for policymakers and public institutions to make necessary short-term decisions. The paper proposes data recording and machine-learning-based analysis using data to validate country forecasts in order to predict COVID-19 proliferation trends and assess risk. The authors are thankful for the high-performance computing resources provided by the Information Technology Open Access Center at the Faculty of Mathematics and Informatics of Vilnius University Information Technology Research Center. Read article.

The scientists of
Institute of Data Science and Digital Technologies and
Institute of Applied Mathematics

Defended dissertation "Development of Tumor Microenvironment-Oriented Digital Pathology Methods for Whole Slide Image Segmentation and Classification" author is thankful for the high-performance computing resources provided by the Information Technology Open Access Center at the Faculty of Mathematics and Informatics of Vilnius University Information Technology Research Center.

Dr. Mindaugas Morkūnas

Increasing intensity in maritime traffic pushes the requirement in better prevention-oriented incident management system. The scientists research results depict that both the autoencoder-based and wild bootstrapping region prediction algorithms can predict vessel trajectory and be applied for abnormal marine traffic detection by evaluating obtained prediction region in an unsupervised manner with desired prediction probability. Information about the research can be found in this article https://www.journals.vu.lt/nonlinear-analysis/article/view/23056/23620. Researchers thankful for the Information Technology Open Access Center at the Faculty of Mathematics and Informatics of Vilnius University for the supercomputer resources provided during this research.

The scientists of
Institute of Data Science and Digital Technologies and
Institute of Applied Mathematics

Defended dissertation "Semi-supervised and Unsupervised Machine Learning Methods for Sea Traffic Anomaly Detection" author is thankful for the high-performance computing resources provided by the Information Technology Open Access Center at the Faculty of Mathematics and Informatics of Vilnius University Information Technology Research Center.

Dr. Julius Venskus

"In this research, a study of cross-linguistic speech emotion recognition is performed. For this purpose, emotional data of different languages (English, Lithuanian, German, Spanish, Serbian, and Polish) are collected, resulting in a cross-linguistic speech emotion dataset with the size of more than 10.000 emotional utterances. Despite the bi-modal character of the databases gathered, our focus is on the acoustic representation only. The assumption is that the speech audio signal carries sufficient emotional information to detect and retrieve it. Several two-dimensional acoustic feature spaces, such as cochleagrams, spectrograms, mel-cepstrograms, and fractal dimension-based space, are employed as the representations of speech emotional features. A convolutional neural network (CNN) is used as a classifier. The results show the superiority of cochleagrams over other feature spaces utilized. In the CNN-based speaker-independent cross-linguistic speech emotion recognition (SER) experiment, the accuracy of over 90% is achieved, which is close to the monolingual case of SER. Read article "

Institute of Data Science and Digital Technologies
Image and Signal Analysis Group

Glaucoma is the second eye disease causing blindness worldwide. Optic Cup-to-Disc ratio (CDR) is a commonly applied method in glaucoma detection. The CDR is calculated based on Optic Disc (OD) and Optic Cup (OC) in eye fundus image screening. Therefore, the accurate segmentation of these two parameters is very important. Lately, Deep Neural Networks have demonstrated great effort in automated Optic Disc and Optic Cup segmentation but the overlapping between regions of OC and OD cause the challenge to obtain CDR automatically with high accuracy. In this paper, we assess the performance of CDR evaluation on three modifications of the Convolutional Neural Network (CNN) U-Net, namely Attention U-Net, Residual Attention U-Net (RAUNet), and U-Net++ applied on publicly available datasets RIM-ONE, DRISHTI, and REFUGE. We calculated the ground truth CDR value of testing eye fundus images of these datasets and compared it with the CDR value obtained by trained CNNs. Our results show that Attention U-net obtains the closest CDR to the ground truth CDR value but the identification of early-stage glaucoma needs an improvement. The authors of the artice "Deep Neural Networks application for Cup-to-Disc ratio estimation in eye fundus images" would like to thank the IT Research Center of Vilnius University for the provided HPC resources.

S. Virbukaitė

"At the Institute of Data Science and Digital Technologies, VU MIF's HPC service infrastructure was used for doctoral studies. The goal of my dissertation research is to investigate deep recurrent neural networks for monitoring the state of sea traffic - predicting the ship's progress. VU MIF's supercomputer resources with GPU calculations were used to create artificial intelligence-based models that solve the regression task. The models created from historical big AIS (Automatic Identification System) data are able to predict the ship's trajectory into the future and this allows to assess the collision risks of ship traffic and detect anomalies at sea. As artificial neural networks are not deterministic, the experiments were repeated several times to ensure the robustness of the results, which was well achieved with the available MIF HPC resources, which can be easily accessed and computed. A great recognition of all this work is the publication of a scientific paper in a highly-ranked EAAI journal, in which the results of the study are presented."

R. Jurkus

“We used the VU MIF supercomputer for large-scale computations to model chemical processes in electrochemical microscopy. The results have been published in high-level application journals. "

R. Astrauskas

“MIF supercomputer resources are useful in interdisciplinary research with scientists at the VU GMC Institute of Biochemistry. The subject of these investigations is the electrochemical properties of damaged (defective) phospholipid membranes. The supercomputer analyzes a large number of three-dimensional models of these membranes using the finite element method (FEM). This simulation avoids the complexity of working with real membranes in the laboratory and allows you to obtain the required information on the properties of the membranes being studied more quickly."

T. Raila

“At VU MIF Supercomputer, we are conducting research on algorithms for electroencephalogram (EEG) analysis. A supercomputer is especially useful for such research because we can test many algorithm configurations at a time and further explore the best one. Other algorithms we can parallelize, then we get acceleration on the supercomputer. We are currently investigating algorithms for EEG peak search and classification based on diagnosis, and apply the results to writing a doctoral dissertation, preparing scientific publications, and giving presentations at international and national scientific conferences. ”

A.V. Misiukas Misiūnas

The graduate of the VU Faculty of Medicine is thankful for the provided IT resources by the ITAPC team. She managed to defend her master's thesis on the topic "Investigation of eye fundus image quality on vascular segmentation using deep neural networks" successfully.

Successfully defended bachelor thesis "An investigation of semi-supervised training methods for facial feature prediction" had been written by the author, who has graduated the "Information Systems Engineering" bachelor's studies. He is grateful for the opportunity to use ITAPC supercomputer for study purposes. The aim of the thesis research was to analyse and create semi-supervised learning method which would produce valuable information about unlabelled data and help during the training phase.

The graduate, who has successfully completed the VU MIF bachelor's study program "Econometrics" and defended her bachelor's thesis on the topic "Assessing the popularity of youtube videos using publicly available metadata", thanks the Information Technology Open Access Research Center for resources. This paper analyses the video metadata and derived parameters collected by the YouTube Data Application Programming Interface. After assessing significant differences in means among popularity groups by using statistical tests, data clustering algorithms: k-means and self-organising neural networks were applied, to search for trends in the formation of popularity groups. Classification of video popularity was also performed using three machine learning methods: support vector machine, random forest and multinomial logistic regression classifiers. A comparative analysis of the algorithms was carried out and the feature sets that most accurately classify the popularity level were selected. The aim of the final work was to contribute to the development of exception identification algorithms that are able to efficiently process traffic data.

The graduate, who has successfully completed the VU MIF master's study program "Modelling and Data Analysis" and defended his master's thesis on the topic "Outlier detection in multidimensional streaming data", thanks the Information Technology Open Access Research Center. The aim of the final work was to contribute to the development of exception identification algorithms that are able to efficiently process traffic data. VU MIF IT OARC supercomputer resources were used for the research of the master's thesis.

The company ZIVE has developed an electrocardiogram device that can be worn on a person's chest and used at home during sports or any other daily activities. Heart activity can be monitored 24x7. When connected to a smartphone or tablet, the app displays detailed heart activity data in real time. After analyzing the data with an AI model, early diagnosis of heart disease can be made, thus providing added value to the user. With the help of MIF scientists and usage of VU MIF HPC resources the AI algorithm was extended with new parameters and their impact on the results was investigated. The developed AI solution is being deployed in medical technology, in collaboration between business and science, using the latest high-performance computing techniques to bring early cardiac diagnostics to the user. Read more

Company "Zivė"

At the Faculty of Mathematics and Informatics of Vilnius University, with the help of HPC resources, Novian Technologies performed the design of another supercomputer for solving the problems of hydrometeorology in one of the 10 largest countries in the world. Based on the computational tasks provided by the customer, Novian Technologies has designed an individual HPC infrastructure (computing nodes, storage, network bandwidth, etc.) capable of solving specific hydrometeorological / climate change challenges in a given time. We are pleased to be able to present the customer with highly accurate results from calculations performed in a real-world HPC environment and using real-world computing power rather than assumptions. Accurate calculation results allowed the customer to safely plan the required capacity of equipment without a large margin, and thus save money.

Ltd. Novian Technologies

“The IT Research center team is characterized by promptness and goodwill. Effective communication allows for productive time planning, which is especially important during events, and goodwill collaboration ensures the quality of our organization's goals. ”

PI "Turing School"

"It's easy to work with the IT Research center team. There are rules in mature business where a verbal agreement equates to a written commitment. This way of communicating allows results to be achieved very quickly without bursting into the bureaucratic jungle."

Ltd. Nevda