Journal of Computer and Forensic Sciences

An Overview of Image Processing in Biomedicine Using U-Net Convolutional Neural Network Architecture

Aleksa Komosar — 2024-02-22

Image processing in biomedicine is a very broad field, which includes both medical and technical significance. The aim of this work is to investigate the current trends in the domain of application of U-Net architecture in the period from 2018 to 2023. The PRISMA framework was used for the systematic literature review and 4 research questions were asked. For the most part, U-Net architectures are used that can process complex high-resolution images in the fastest way in the context of semantic segmentation. Previous work in image processing has focused on overcoming problems such as the complexity of different architectures, image loss, image resolution, and quality, as well as the size of datasets and noise reduction. The most frequently used groups of datasets are BraTS, Data Science Bowl, and ISIC Challenge. The best general Dice score was obtained for LUNA16, VESSEL12, and The Kaggle Lung datasets with 0.98. It is concluded that the application of the U-net network is growing, with a focus on solving specific challenges in the context of a certain modality and segment of biomedicine.

An Approach to Optimization of Gated Recurrent Unit with Greedy Algorithm

Patricio Ignacio Lazcano Muñoz — 2024-02-27

This study focuses on enhancing the performance of Stacked Gated Recurrent Unit (GRU) model in time series data processing, specifically in stock price prediction. The most significant innovation occurs in the integration of a Greedy Algorithm for optimizing hyperparameters such as look-back period, number of epochs, batch size, and units in each GRU layer. Historical stock data from Apple Inc. is utilized for the model's application, emphasizing the model's effectiveness in predicting stock prices. The study methodology involves a sequence of steps, such as data loading, preprocessing, dataset splitting, model construction and evaluation. The role of the Greedy Algorithm's focuses in iteratively adjusting hyperparameters to minimize the Root Mean Squared Error (RMSE) metric, resulting in refining the model's predictive accuracy. The outcomes reveal that the integrated Greedy Algorithm significantly enhances the model's accuracy in predicting stock prices, indicating its potential application in various scenarios requiring precise time series forecasting.

Merging Control-flow and Dataflow Architectures on a Single Chip

Nenad Korolija — 2024-04-10

Computing power rises predominantly by increasing the number of cores of modern processors, and the number of cores in cluster and cloud architectures. Along with increasing the processing power, high performance computing requirements also rise. Vast of the computing infrastructure includes control-flow processors that are based on von Neumann paradigm. In contrary, the principle of dataflow architectures is based on the data flowing through the already configured hardware. Dataflow architectures are often implemented using FPGAs. Recent research has proposed hybrid architectures, where both control-flow and dataflow hardware would exist on the same chip die. In this article, a new hybrid control-flow and dataflow architecture is proposed, where control-flow hardware would be similar to modern graphical cards, consisting of thousands of cores, but with a reasonable small amount of dataflow hardware available on each GPU core. The proposed architecture is tested by analyzing the conjugate gradient method executed on both control-flow and dataflow hardware. The execution of a the algorithm is divided onto GPU cores, and the execution of repeated instructions on each GPU core is delegated to the assigned dataflow hardware. Results indicate that it is possible to accelerate the execution of algorithms using the proposed architecture.

The New Language Corpus: Exploring Perception of Graphemes 'P' and 'B' in Serbian Cyrillic and Latin Scripts According to Semantical Context

Sara Tvrdišić — 2024-05-20

Language corpus serve as essential instruments in facilitating research endeavors within the domain of natural language processing. As an illustration of such a language corpus, the paper presents part of the conducted research. Notably, the Serbian language demonstrates a deficiency in its linguistic resources, a concern accentuated by its unique status as one of the few languages characterized by a bialphabetical system., i.e. although in official use only the Cyrillic alphabet is used, in the Serbian language the Latin alphabet is also used in parallel. The Serbian Cyrillic alphabet in modern times is threatened by various influences of globalization, i.e. more dominant use of the Latin alphabet in everyday life. The primary goal of the research is to examine the perceptual differences between the Latin and Cyrillic alphabets (for the purposes of the research, a sample of subjects whose native language is not Serbian was used, so that the subjects would not be compelled by previous knowledge of the semantics of the Serbian language - the subjects would have to know at least the basics of the Serbian language and the rules of reading and writing, to be able to understand the questionnaire, simplified with vocabulary for level A2). Within this paper, a segment of this research will be presented, through several compared variables, and predominantly by examining whether the graphemes 'P' and 'B' can cause confusion, because these graphemes in Serbian Cyrillic and Latin have different phonetic pronunciation. The research showed that the answers given by the subjects were in a higher percentage perceived in the reverse script of the questionnaire (if the questionnaire was in Cyrillic, the perception of unclear words by a higher percentage of subjects was in Latin and vice versa).