Light Weight Structure Texture Feature Analysis for Character Recognition Using Progressive Stochastic Learning Algorithm

Advances in computational intelligence and robotics book series, Journal Year: 2023, Volume and Issue: unknown, P. 144 - 158

Published: Dec. 21, 2023

Handwritten character recognition is a challenging task in the field of image processing and pattern recognition. The success systems depends heavily on feature extraction methods used to represent images. In this chapter, authors propose novel method called progressive stochastic learning (PSL) algorithm. proposed work based texture structural features designed extract discriminative that capture essential information characters. PSL algorithm classify extracted into their respective classes. Experimental results demonstrate achieves accuracy 92.6% for correct characters predicted 91.3% words predicted. Moreover, outperforms several state-of-the-art terms accuracy, computation time, memory requirements.

Language: Английский

---

Lung Disease Detection Using Scale-Invariant Weighted Ensemble Neural Architecture

Lecture notes in networks and systems, Journal Year: 2025, Volume and Issue: unknown, P. 83 - 94

Published: Jan. 1, 2025

Language: Английский

---

COVID-19 Detection from Optimized Features of Breathing Audio Signals Using Explainable Ensemble Machine Learning

Results in Control and Optimization, Journal Year: 2025, Volume and Issue: unknown, P. 100538 - 100538

Published: Feb. 1, 2025

Language: Английский

---

Graphical Insight: Revolutionizing Seizure Detection with EEG Representation

Biomedicines, Journal Year: 2024, Volume and Issue: 12(6), P. 1283 - 1283

Published: June 10, 2024

Epilepsy is characterized by recurring seizures that result from abnormal electrical activity in the brain. These manifest as various symptoms including muscle contractions and loss of consciousness. The challenging task detecting epileptic involves classifying electroencephalography (EEG) signals into ictal (seizure) interictal (non-seizure) classes. This classification crucial because it distinguishes between states seizure seizure-free periods patients with epilepsy. Our study presents an innovative approach for neurological diseases using EEG leveraging graph neural networks. method effectively addresses data processing challenges. We construct a representation extracting features such frequency-based, statistical-based, Daubechies wavelet transform features. allows potential differentiation non-seizure through visual inspection extracted To enhance detection accuracy, we employ two models: one combining convolutional network (GCN) long short-term memory (LSTM) other GCN balanced random forest (BRF). experimental results reveal both models significantly improve surpassing previous methods. Despite simplifying our reducing channels, research reveals consistent performance, showing significant advancement neurodegenerative disease detection. accurately identify signals, underscoring streamlined not only maintains effectiveness fewer channels but also offers visually distinguishable discerning opens avenues analysis, emphasizing impact representations advancing understanding diseases.

Language: Английский

---

Technological Frontier on Hybrid Deep Learning Paradigm for Global Air Quality Intelligence

Advances in computational intelligence and robotics book series, Journal Year: 2024, Volume and Issue: unknown, P. 144 - 162

Published: May 31, 2024

This hybrid deep-learning study focuses on pollutant concentration. It illuminates convolutional neural networks (CNN) and long short-term memory in deep learning methods (LSTM). CNNs are essential to learning, especially image processing. They ideal for pollution concentration analysis because they extract complex data features. LSTM is another important tool this study. LSTMs recurrent (RNNs) that can process store sequences. Time-series analysis, common research, benefits from them. Understanding learning's impact issues. investigates a CNN-LSTM model combines CNN feature extraction with sequence fusion lets the make smart predictions input PCA key investigation. dimensionality reduction finds variables significant relationships.

Language: Английский

---

Creating a Sustainable Large-Scale Content-Based Biomedical Article Classifier Using BERT

Advances in computational intelligence and robotics book series, Journal Year: 2024, Volume and Issue: unknown, P. 290 - 303

Published: May 31, 2024

Given the scarcity of labeled corpora and high costs human annotation by qualified experts, clinical decision-making algorithms in biomedical text classification require a significant number costly training texts. To reduce labeling expenses, it is common practice to use active learning (AL) approach volume documents required produce performance. There are two methods for categorizing articles: article-level journal-level classification. In this chapter, authors present hybrid strategy classifiers with article metadata such as title, abstract, keywords annotated FoR (fields research) using natural language processing (NLP) embedding techniques. These then applied at level analyze publications PubMed metadata. The trained BERT codes them classify based on their available

Language: Английский

---

Insights Into AI Systems for Recognizing Human Emotions, Actions, and Gestures

Advances in computational intelligence and robotics book series, Journal Year: 2024, Volume and Issue: unknown, P. 389 - 410

Published: Aug. 28, 2024

The capacity of AI systems to recognize and interpret human emotions, actions, gestures is reshaping numerous sectors, from entertainment security. This chapter offers a comprehensive review the current state-of-the-art technologies in this domain, shedding light on their strengths, potential limitations, avenues for improvement. In recent years, capable recognizing understanding have shown remarkable progress. They are deployed diverse applications, including virtual reality, healthcare, human-computer interaction. However, capabilities limitations crucial harness fully. Our research provides deep dive into strengths existing systems, showcasing ability accurately decipher complex expressions, movements, emotional states. We also critically examine such as bias training data or challenges subtle cultural nuances.

Language: Английский

---

Predicting Child Mortality With Diverse Regression Algorithms Using a Machine Learning Approach

Advances in computer and electrical engineering book series, Journal Year: 2024, Volume and Issue: unknown, P. 329 - 352

Published: Aug. 29, 2024

This chapter uses machine learning methodologies to investigate the prediction of child mortality rates for ages 1-4 across diverse countries. Drawing upon a comprehensive review global health data from organizations such as World Health Organization (WHO) and United Nations Children's Fund (UNICEF), which highlight urgency significance accurate prediction, authors analyze dataset spanning 1967 2019, containing 30,940 entries countries worldwide. Regression algorithms, including XGBoost, CatBoost, Random Forest, AdaBoost, DecisionTree Regressor, are employed predict rates. Evaluation metrics R^2, adjusted mean absolute error (MAE), squared (MSE), root (RMSE) utilized assess model performance. Additionally, Matplotlib Seaborn use visualization techniques illustrate findings through pie charts graphs. The analysis aims identify most effective algorithm accurately forecasting rates, thereby contributing advancing healthcare planning intervention strategies reduce globally.

Language: Английский

---

Graph Neural Networks for Neurological Diseases Detection in EEG Signals

Research Square (Research Square), Journal Year: 2023, Volume and Issue: unknown

Published: July 6, 2023

Abstract Epilepsy is a neurological disorder that affects large portion of the world's population1. It characterized by recurrent seizures2, which are caused abnormal electrical activity in brain3. This condition can lead to various symptoms, such as muscle contractions, loss consciousness, and convulsions. Here, we present novel approach for detecting seizures other diseases electroencephalogram (EEG) signals highlight potential graph neural networks EEG signal analysis, suggesting this method could effectively tackle challenges related data processing analysis. The involves constructing representation extracting frequency-based, statistical-based, Daubechies wavelet transform features from signals, using sliding window with 30% overlap. A Graph Convolutional Network (GCN) model, combined Long Short-Term Memory (LSTM) architecture (GCN-LSTM) Balanced Random Forest (GCN-BRF), then employed process improve accuracy seizure detection. Our experimental findings show both GCN-LSTM GCN-BRF models significantly improved detection, surpassing performance previously published methods. Despite reducing complexity our decreasing number channels used, observed its remains unchanged. means even fewer channels, continues deliver comparable results without compromising effectiveness. We anticipate work be significant step forward field neurodegenerative diseases, illustrating ability accurately detect signals. promising providing new avenue future research. Furthermore, utilization offers clear visually distinguishable differentiate normal

Language: Английский

---

Binary modified Cat and Mouse based Optimizer for medical feature selection: A COVID-19 case study

Research Square (Research Square), Journal Year: 2024, Volume and Issue: unknown

Published: March 13, 2024

Abstract Recent technological advances in medical diagnosis have led to the generation of high-dimensional datasets. The presence redundant and irrelevant features these datasets can adverse effects on performance machine learning (ML) methods reduce accuracy their results. Therefore, feature selection (FS), i.e., a popular preprocessing method ML, is used select optimal subsets improve ML methods. This enhancement more crucial while addressing issues. Since FS multiobjective binary optimization problem, it necessary develop efficient algorithms. Although metaheuristic algorithms (MAs) been widely for medicine, they face different challenges most applications, e.g., lack sufficient effectiveness scalability effective small large cat mouse-based optimizer (CMBO) novel MA based natural competitive behavior cats mice. Despite its acceptable variety problems, CMBO faces various such as limited exploitation abilities, an unbalanced search mechanism, high fluctuation solutions complex FS. paper proposes modified version called BMCMBO enhance selecting from involves significant modifications updating positions agents, mice, effect positional information member population, addition adaptive step size. These are meant boost solutions, balance process when dealing with problem proposed algorithm 12 real was compared variants. statistical results demonstrated that than other evaluated In addition, employed diagnose COVID-19 case study. identified healthy infected correctly samples 98.4\%, demonstrating superiority.

Language: Английский

---