Topic title |
Possible scientific supervisors |
Source of funding |
Application of Generative Adversarial Networks for novelty detection in one-class problems
|
doc. dr. Arūnas Lipnickas |
state-funded |
Research Topic Summary.
One-class novelty detection is a serious problem that receives relatively little attention in the debate about machine learning problems.
Generative Adversarial Networks (GAN) networks are finding more and more of their own applications because they need less input data for training compared to artificial-intelligence-based systems. The purpose of this work is to determine the capabilities of the GAN network to recognize the defect of the textured surface pattern in a limited number of samples of defect training.
|
A study on generative methods for creating virtual reality learning objects |
prof. dr. Tomas Blažauskas |
state-funded |
A method to control glycaemia in intensive care and predict risks
|
prof. dr. Rytis Maskeliūnas |
state-funded |
Research Topic Summary.
The aim of this topic is to develop an artificial intelligence-based system for glycaemic control in patients treated at the LSMU Intensive Care Clinic, in order to optimise glycaemic control and to predict the risks associated with insulin dosage. The main objective is to develop an method that uses glycaemic measurements and other clinical data to automatically determine the optimal amount of insulin to maintain glycaemia below 10 mmol/l. The method would potentially "learn" from the patient's previous responses to insulin doses, personalising the therapeutic plan. In addition, the AI would be trained to predict the risk of hypoglycaemia, ensuring that glycaemic control is effective while minimising the risk of hypoglycaemia. Such an application would not only improve the quality of life and outcome of patients, but also reduce healthcare costs by optimising glycaemic control.
|
Methods for classification of uncertain shape fast moving objects
|
doc. dr. Armantas Ostreika |
state-funded |
Research Topic Summary.
When applying artificial intelligence (AI) algorithms, it is important to properly select criteria that combine the essential properties of the object or phenomenon under consideration, and thus to create an appropriate criteria/parameter model that allows unambiguous identification and classification of object state changes. In computer vision tasks, the creation of such models is not a trivial task, as visual information is enormously diverse and, moreover, changes dynamically.
Properly designed and implemented, such models serve for high-level decision-making using a digital image or video data. These solutions are designed to automate tasks that can only be performed by a human. Computer vision methods allow automatic analysis of important information in the processing of large data flows. AI algorithms in this case are an integral part of solving computer vision problems.
The proposed topic will aim to improve existing common methodologies by focusing on specific products and real-world processes for greater efficiency and reliability, adapting and improving digital image acquisition, processing, analysis and interpretation methods, machine learning and AI methods including convolutional neural networks, hybrid convolutional network models, real-time object detection algorithms (YOLO), methods supported by autoregressive language models (GPT-3) etc.
Focusing on real examples of provided services or production processes, the aim will be to apply and test the developed models and methodologies in practice. Examples of applications include computer vision systems that measure and count products or blanks, estimate their weight, shape, or volume, and inspect objects at high speed based on their predefined characteristics.
In this research, we expect to overcome operative restrictions that hamper the adoption of new algorithms. It is expected to modify existing methodologies and to implement novel research ideas based on AI and new heuristic approaches.
|
Development of hybrid optimization method for AI infrastructure using functionals
|
prof. dr. Renaldas Urniežius |
state-funded |
Research Topic Summary.
In this topic, the Ph.D. student will develop hybrid methods for AI infrastructure optimization (e.g., Mobile NET-SSD, Yolo, and others), which will improve the efficiency of the existing infrastructure in terms of training, speed, and the number of infrastructure parameters. Tasks will include the development of a hybrid method, a fusion solution using functionals, and practical validation with a selected practical system.
|
Development and effectiveness analysis of smart systems for information integration and healthcare
|
prof. dr. Robertas Alzbutas |
state-funded |
Research Topic Summary.
The objective of the research is the creation of a methodology for the effective application of information integration and smart systems intended for health care.
Tasks:
1. Review and compare possibilities of information integration methods and algorithms as well as the application of related software and smart systems intended for health care.
2. Define information integration and smart systems efficiency criteria and their evaluation procedures in the context of data-relevant health care.
3. Develop and demonstrate the information integration tools which usage increase the speed and/or reduce the usage of computing resources.
4. Perform the pilot studies of smart systems intended for health care and create a methodology for these systems’ effective application.
This research will allow the development of such modeling and data analytics tools and their application methodology, which will enable faster data synthesis, less resource-intensive decision-making, and more effective health care intended smart systems (e.g. smart clothing used for health monitoring) application.
For further information please contact the supervisor of the topic.
|
Signal processing and machine learning for monitoring and risk analysis of battery heterogenic ageing
|
prof. dr. Robertas Alzbutas |
state-funded |
Research Topic Summary.
More accurate estimation of battery State of Charge (SOC) and State of Health (SOH), as well as optimization of charging cycles and overall battery usage, could benefit from decision support tools that better (faster and more precise) utilize and translate the information held in signals coming from Battery-Management System (BMS).
The objective of the research is to develop a method and system, used to support BMS, able to autonomously monitor battery SOH and estimate the risk of different types of malfunctions and which will provide easy-to-interpret recommendations and decisions for battery usage.
Tasks and expected results:
1. Data collection and development of deep neural network architectures for SOH monitoring (identification, characterization, and classification) and dealing with data heterogeneity as well as time dependence.
2. Application and testing of developed models and risk analysis approach for the case of various data (e.g. sample of annotated BMS data), taking into consideration data transformation optimization and calibration results.
3. Software development, for deep learning of SOH monitoring, when taking into consideration data heterogeneity and risk analysis. The software will be released under the GNU General Public License for the free usage of interested parties.
For further information on the topic, and relevant R&D work please contact the supervisor.
|
Enhancing Brain-Computer Interface (BCI) Systems through Hybrid EEG and Virtual Reality (VR) Platforms
|
prof. dr. Robertas Damaševičius |
state-funded |
Research Topic Summary.
This PhD research project aims integrating Electroencephalography (EEG) and Virtual Reality (VR) into a Brain-Computer Interface (BCI) system for monitoring and medical rehabilitation of patients. By developing robust algorithms for interpreting EEG data the project aims for better understanding and utilization of brain signals in various applications.
|
Enhancing Brain Tumor Recognition and MRI Segmentation Through Vision Transformers and Explainable AI
|
prof. dr. Robertas Damaševičius |
state-funded |
Research Topic Summary.
This topic outlines a comprehensive study on leveraging Vision Transformers (ViTs) for brain tumor recognition and MRI segmentation. The topic involves designing Vision Transformers specifically for brain tumor recognition and MRI segmentation, evaluating their performance against current deep learning methods, and assessing their generalization across diverse datasets. The research aims to use explainable AI techniques to interpret the decision-making process of ViTs. This includes developing AI visualization methods to display learned features and decision pathways.
|
A study on the use of virtual reality and tactile technologies for learning and control |
prof. dr. Tomas Blažauskas |
state-funded |
A low-resolution forest point cloud segmentation method for forest inventory
|
prof. dr. Rytis Maskeliūnas |
state-funded |
Research Topic Summary.
This research endeavors to pioneer a revolutionary low-resolution forest point cloud segmentation method for forest inventory, aiming to surpass existing methodologies by seamlessly integrating state-of-the-art technologies. The primary objective is to develop a robust and adaptive algorithm that effectively addresses the challenges posed by low-resolution point cloud data, enabling precise identification and segmentation of distinct elements within complex forest environments. Through the incorporation of deep (machine) learning, LiDAR, and potential data fusion with multispectral imaging, the proposed topic seeks not only to enhance the accuracy of tree identification and biomass estimation but also to potentially facilitate comprehensive forest management. The research further aims to optimize computational efficiency, ensuring applicability across large-scale forested areas. By achieving these objectives, this groundbreaking approach aspires to set a new benchmark in forest inventory methodologies, fostering sustainable forestry practices and contributing to our understanding of forest ecosystems on a broader scale.
|
Human emotion recognition using a hybrid Brain-Computer Interface (BCI) and Muscle-Computer Interface (MCI) system
|
prof. dr. Robertas Damaševičius |
state-funded |
Research Topic Summary.
This PhD topic is offered by the European Doctoral Network for Neural Prostheses and Brain Research (DONUT). The aim of the topic is to develop and explore the methods necessary for the recognition of externally expressive emotions (positive, neutral, negative) by using a hybrid neuronal (EEG and EMG) interface.
More information about the topic, admission requirements and selection process is available at https://euraxess.ec.europa.eu/jobs/186013
The length of doctoral studies under this topic is three years. Start of studies is 1st September, 2024.
Candidates to this topic shall meet eligibility criteria specified in the offer and undergo two stages of selection process:
1 stage. Send all required documents to dc9@donut-project.eu till 31 March, 2024.
2 stage. The candidates selected by a project’s recruitment committee shall apply to KTU and upload all necessary documents following the confirmed deadlines
|