3 sonuçlar
Arama Sonuçları
Listeleniyor 1 - 3 / 3
Yayın ANN activation function estimators for homomorphic encrypted inference(Institute of Electrical and Electronics Engineers Inc., 2025-06-13) Harb, Mhd Raja Abou; Çeliktaş, BarışHomomorphic Encryption (HE) enables secure computations on encrypted data, facilitating machine learning inference in sensitive environments such as healthcare and finance. However, efficiently handling non-linear activation functions, specifically Sigmoid and Tanh, remains a significant computational challenge for encrypted inference using Artificial Neural Networks (ANNs). This study introduces a lightweight, ANN-based estimator designed to accurately approximate activation functions under homomorphic encryption. Unlike traditional polynomial and piecewise linear approximations, the proposed ANN estimators achieve superior accuracy with lower computational overhead associated with bootstrapping or high-degree polynomial techniques. These estimators are trained on plaintext data and seamlessly integrated into encrypted inference pipelines, significantly outperforming conventional methods. Experimental evaluations demonstrate notable improvements, with ANN estimators enhancing accuracy by approximately 2% for Sigmoid and up to 73% for Tanh functions, improving F1-scores by approximately 2% for Sigmoid and up to 88% for Tanh, and markedly reducing Mean Square Error (MSE) by up to 96% compared to polynomial approximations. The ANN estimator achieves an accuracy of 97.70% and an AUC of 0.9997 when integrated into a CNN architecture on the MNIST dataset, and an accuracy of 85.25% with an AUC of 0.9459 on the UCI Heart Disease dataset during ciphertext inference. These results underscore the estimator’s practical effectiveness and computational feasibility, making it suitable for secure and efficient ANN inference in encrypted environments.Yayın Supervised decision making in forex investment using ML and DL classification methods(Işık Üniversitesi, 2023-07-20) Jiroudi, Abdullah; Eskil, Mustata Taner; Işık Üniversitesi, Lisansüstü Eğitim Enstitüsü, Bilgisayar Mühendisliği Yüksek Lisans Programı; Işık University, School of Graduate Studies, Master’s Program in Computer EngineeringThe suggested trading system offers an approach that takes into account the complexity and high trading volume of the foreign exchange (FX0) market. Its main objective is to address the challenges faced by traders in the GBP/JPY currency pair and assist them in making quick decisions. To achieve this, machine learning and deep learning techniques are integrated to propose a trading algorithm. The proposed algorithm works by combining data from different time intervals. The Long Short-Term Memory (LSTM) model is used to predict indicator values, while the XGBoost classifier is employed to determine trading decisions. This method aims to adapt to rapidly changing patterns in the forex market and enables the detection of subtle changes in price dynamics through a sliding window training approach. Experiments conducted have shown promising results for the suggested trading system. Positive outcomes have been obtained in terms of capital growth and prediction accuracy. However, since this method is highly risky and requires further development in terms of risk management, the inclusion of risk management techniques and algorithm optimization is targeted. This study contributes to the improvement of trading strategies while bridging the gap between researchers and traders. It also demonstrates the potential of machine learning and deep learning techniques to enhance decision-making processes in financial markets. This trading system offers traders a range of advantages. The utilization of machine learning and deep learning techniques enables rapid analysis of large amounts of data and decision-making capabilities. Additionally, by combining data from different time intervals, it becomes possible to evaluate long-term trends and short-term fluctuations more effectively. In conclusion, the suggested trading system empowers traders to be competitive in the forex market and achieve better outcomes. Furthermore, it contributes to the increased utilization of machine learning and deep learning techniques in financial markets and encourages further research in the field.Yayın Theta and Beta1 frequency band values predict dyslexia classification(John Wiley and Sons Ltd, 2025-12-29) Eroğlu, Günet; Harb, Mhd Raja AbouDyslexia, impacting children's reading skills, prompts families to seek cost-effective neurofeedback therapy solutions. Utilising machine learning, we identified predictive factors for dyslexia classification. Employing advanced techniques, we gathered 14-channel Quantitative Electroencephalography (QEEG) data from 200 participants, achieving 99.6% dyslexic classification accuracy through cross-validation. During validation, 48% of dyslexic children's sessions were consistently classified as normal, with a 95% confidence interval of 47.31 to 48.68. Focusing on individuals consistently diagnosed with dyslexia during therapy, we found that dyslexic individuals exhibited higher theta values and lower beta1 values compared to typically developing children. This study pioneers machine learning in predicting dyslexia classification factors, offering valuable insights for families considering neurofeedback therapy investment.
