Presentation materials
I. BACKGROUND
Pulmonary Embolism (PE) is reported to be one of the most
common cardiovascular diseases. It is caused by a blood clot
that develops in a blood vessel elsewhere in the body and
travels to an artery in the lung, forming a blockage [1].
Lung V/Q (ventilation/perfusion) SPECT is one of an es-
tablished diagnostic imaging test for suspected PE. The idea
behind this test is...
Recent advances in deep learning and natural language processing have spurred the demand for deploying increasingly complex models on resource-constrained platforms. Modern browser environments, empowered by emerging GPU standards like WebGPU, now offer a promising venue for real-time AI inference. This paper provides an overview of leveraging WebGPU for accelerating inference directly within...
Missing values are a common phenomenon in real-world time series datasets and can significantly impact the precision and reliability of data analysis and machine learning models. This research project aims to discuss the types of missing data occurrence and test and analyze different possibilities of their imputation. The methods taken into consideration will start from the simplest ones based...
Japanese manga has captivated readers worldwide with its vibrant and expressive form of art that provides compelling storytelling with intricate visuals. However, for many fans outside of Japan, language barriers often stand in the way of fully experiencing the depth of these stories. Traditional translation from language to language, while effective, can be a very time-consuming and...
This paper investigates short-term energy produc-
tion forecasting in Poland using three distinct predictive mod-
els: LightGBM, an ARIMA-based model, and a Long Short-
Term Memory (LSTM) network. Leveraging historical data
from Polandโs energy sector, our study evaluates each modelโs
performance in terms of accuracy, robustness, and computational
efficiency. The LightGBM model employs...
Large Language Models have gained widespread
recognition since OpenAI released their revolutionary model,
ChatGPT 3.5. Since then, many new approaches have emerged
to improve the capabilities and accuracy of these models for
different tasks. One such method involves using multi-agent
conversations. This article compares two multi-agent setups
designed to solve the Polish standardized...
Automated medical report generation from chest X-ray images is a critical area of research in medical AI, aiming to enhance diagnostic accuracy, reduce radiologists' workload, and improve patient care. The process involves analyzing medical images and translating visual findings into structured, clinically relevant textual reports. Traditional methods rely on human expertise, which is...
This article investigates the role of Retrieval-Augmented Generation (RAG) in enhancing Large Language Models (LLMs) with information about movies and TV series released beyond their training data. In this study, the Llama 3.2 3B LLM is leveraged and integrated with external movie-related data retrieved from the OMDb API to provide specific information about over 14000 titles released in 2024,...
With the increasing popularity of web services and the significant amount of time users spend online, cybercriminals are increasingly targeting this space with sophisticated malicious techniques. Phishing, one of the most prevalent cybersecurity threats, poses a substantial risk to both individuals and organizations. These attacks are typically executed via deceptive emails containing...
Short-term forecasting of cryptocurrency prices remains a challenging task due to the high volatility and complex market dynamics of digital assets like Ethereum. This study proposes a hybrid deep learning model, that integrates networks such as Bi-LSTM, FinBERT and GRU, seeking to provide a comprehensive analysis of their applicability in this domain and enhance predictive accuracy for this...
The paper aims to evaluate the effectiveness of available detectors in distinguishing fake from real photos. Deepfakes are artificially generated images or videos created using artificial intelligence or manual tools like Photoshop. New techniques, such as Generative Adversarial Networks (GANs) and Diffusion Models (DMs) enable the rapid generation of highly realistic images. The research...
Modern organizations generate vast amounts of data, a significant portion of which consists of system, network, and application logs. The sheer volume and scalability of these logs make manual analysis inefficient and highly time-consuming. Automated anomaly detection techniques have been developed and refined to accelerate this process. Currently it is common for open SIEM systems to only...
This paper investigates the utility of Principal Component Analysis (PCA) for multi-label classification of multispectral images using ResNet50 and DINOv2, acknowledging the high dimensionality of such data and the associated processing challenges. Multi-label classification, where each image may belong to multiple classes, adds further complexity to feature extraction. Our pipeline includes...
Understanding and interpreting the decisions made by deep learning models has become an essential area of research in artificial intelligence. Convolutional neural networks (CNNs), despite their high performance in various tasks, often function as "black boxes," making it challenging to explain their predictions. This study focuses on applying and evaluating different explainability techniques...
Large language models (LLMs) are trained with ever-increasing amounts of data. It seems that when asked to solve mathematical tasks, they can infer and reason mathematically [1]. The GSM8K benchmark platform is widely used to test various LLMs to solve simple arithmetic tasks. Recently, LLMs have shown a clear improvement in their ability to correctly answer questions from the GSM8K dataset....
Franky: An Intelligent Agent for Stock Portfolio
Management Using Large Language Models and
Deep Reinforcement Learning
1st Mikoลaj Zawadaโ
Faculty of Electrical Engineering
Warsaw University of Technology
Warsaw, Poland
mikolaj.zawada.stud@pw.edu.pl
2nd Mateusz Bartosikโ
Faculty of Electrical Engineering
Warsaw University of Technology
Warsaw,...
This paper presents a hybrid framework combining Deep Neural Networks (DNNs) and Modern Portfolio Theory (MPT) to optimize football betting strategies. Leveraging historical match data from five major European leagues (2014โ2025), we engineer predictive features such as dynamic Elo rankings, expected goals (xG), and team performance metrics derived from raw game statistics. Diverse neural...
This paper presents a comprehensive approach to the evaluation of physical exercise using skeletal estimation and motion analysis. The research begins with the acquisition of relevant datasets and a review of existing monocular pose estimation solutions. A key component of this work is modeling important features, including skeletal joint weighting and key angles between them, according to...
The rise of quantum computing presents a significant threat to classical cryptographic systems, particularly those relying on hard mathematical problems such as integer factorization and discrete logarithms. This paper explores the impact of quantum computing on traditional cryptographic algorithms and the necessity of transitioning to quantum-resistant cryptography. We provide an overview of...
This paper presents a comprehensive comparison of several state-of-the-art proxy loss methods for retail product recognition, focusing on accuracy, computational efficiency, embedding quality, and convergence behaviour. An extensive evaluation is performed on the Stanford Online Products benchmark dataset, which contains over 120,000 images covering 22,634 distinct product categories. The...
Sygnaล elektroencefalograficzny (EEG) moลผe ulegaฤ zakลรณceniom spowodowanym przez aktywnoลฤ elektrycznฤ niezwiฤ zanฤ bezpoลrednio z pracฤ mรณzgu badanego. Zjawisko to okreลlane jest jako problem artefaktรณw EEG, ktรณre dzielimy na techniczne i fizjologiczne. Do artefaktรณw fizjologicznych naleลผฤ charakterystyczne fragmenty sygnaลu rejestrowane w wyniku ruchรณw gaลek ocznych, mrugania, czy aktywnoลci...
