About Me
Working with deep learning and deep neural networks since my 2011. My work focuses on deep learning theory and applications, and I’ve contributed to advancements in outliers detection.
Main publications
|
Title: Multispectral Image Caption Unification Using Diffusion and Cycle GAN Models
IEEE Access 2025 Kürşat Kömürcü, Linas Petkevicius Summary:We propose a full-circle generative pipeline that unifies captions, RGB images, and Sentinel-2–like multispectral data by combining zero-shot captioning, fine-tuned Stable Diffusion, and CycleGAN translation to produce credible synthetic triplets from otherwise incomplete geospatial datasets. Read more |
|
Title: Few-Shot Learning for Triplet-Based EV Energy Consumption Estimation Applied Artificial Intelligence 2025 Alminas Čivilis, Linas Petkevicius , Simonas Šaltenis, Kristian Torp, Ieva Markucevičiūtė-Vinckė Summary:This paper introduces a telematics-driven framework for EV travel time and energy prediction—using trajectory triplets and few-shot learning—and shows that the resulting models can robustly adapt to diverse EV types despite drastic contextual changes. Read more |
|
Title: Optimising Quantum Algorithm Schemes According to the Topology of Quantum Computers 2025 IEEE 12th Workshop on Advances in Information, Electronic and Electrical Engineering (AIEEE) Gabrielius Keibas, Linas Petkevicius Summary:This study presents a two-step optimisation method for adapting quantum algorithms to specific quantum computer topologies—qubit mapping followed by SWAP insertion—and shows that while dense, larger topologies improve initial qubit placement, most architectures still require SWAP operations to achieve implementable circuits. Read more |
|
Title: MiniCPM-V LLaMA Model for Image Recognition: A Case Study on Satellite Datasets (IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2025) Kürşat Kömürcü, Linas Petkevicius Summary: This study shows that MiniCPM-V performs promisingly in satellite image recognition across multiple datasets, though with notable class-specific limitations, highlighting LLMs’ potential for advancing remote sensing analysis. Read more |
|
Title: Identification of algal blooms in lakes in the Baltic states using Sentinel-2 data and artificial neural networks (IEEE Access, 2024) Dalia Grendaitė, Linas Petkevicius Summary: This work proposes remote monitoring techniques using satellite images and machine learning algorithms to predict chlorophyll α concentration in water bodies and identify algal blooms. The training and test dataset used in this study includes diverse range of lakes in Baltic countries. Read more |
|
Title: Zero Shot Classification for Change Detection in Satellite Imagery (AIEEE, 2024) Kürşat Kömürcü, Linas Petkevicius Summary: This research investigates the zero-shot classification using the Comparative Language-Image Pre-Training (CLIP) model for change detection in satellite imagery. Read more |
|
Title: Symbolic Neural Architecture Search for Differential Equations (IEEE Access, 2023) Paulius Sasnauskas, Linas Petkevicius Summary: This paper propose the first use of symbolic integration that leverages the machine learning infrastructure, such as automatic differentiation, to find analytical approximations of ordinary and partial differential equations. Read more |
|
Title: Probabilistic Deep Learning for Electric-Vehicle Energy-Use Prediction (SSTD, 2021) Linas Petkevicius, Simonas Saltenis, Alminas Civilis, Kristian Torp Summary: This paper proposes a two-tier architecture using deep learning to predict electric vehicle route travel time and energy use, leveraging EV tracking data and contextual information. It explores various speed profile generation methods and probabilistic deep learning models for energy prediction, validated with real-world datasets. Read more |
|
Title: Topological navigation graph framework (Autonomous Robots, 2021) Povilas Daniušis, Shubham Juneja, Lukas Valatka, Linas Petkevicius Summary: This paper propose topological navigation graph (TNG) framework. TNG is an imitation-learning-based topological navigation framework for navigating through environments with intersecting trajectories. Read more |
|
Title: Multiple Outlier Detection Tests for Parametric Models (Mathematics, 2020) Vilijandas Bagdonavičius, Linas Petkevicius Summary: This paper propose a simple multiple outlier identification method for parametric location-scale and shape-scale models when the number of possible outliers is not specified. Read more |
|
Title: A new multiple outliers identification method in linear regression (Metrika, 2020) Vilijandas Bagdonavičius, Linas Petkevicius Summary: This paper propose new method for multiple outliers identification in linear regression models is developed. The method is based on a result giving asymptotic properties of extreme studentized residuals. Read more |