Mühendislik Fakültesi / Faculty of Engineering
Permanent URI for this collectionhttps://hdl.handle.net/11727/1401
Browse
5 results
Search Results
Item A Systematic Review of Transfer Learning-Based Approaches for Diabetic Retinopathy Detection(2023) Oltu, Burcu; Karaca, Busra Kubra; Erdem, Hamit; Ozgur, Atilla; 0000-0002-9237-8347; 0000-0003-1704-1581; AAD-6546-2019Diabetic retinopathy, which is extreme visual blindness due to diabetes, has become an alarming issue worldwide. Early and accurate detection of DR is necessary to prevent the progression and reduce the risk of blindness. Recently, many approaches for DR detection have been proposed in the literature. Among them, deep neural networks (DNNs), especially Convolutional Neural Network (CNN) models, have become the most offered approach. However, designing and training new CNN architectures from scratch is a troublesome and labor-intensive task, particularly for medical images. Moreover, it requires training tremendous amounts of parameters. Therefore, transfer learning approaches as pre-trained models have become more prevalent in the last few years. Accordingly, in this study, 43 publications based on DNN and Transfer Learning approaches for DR detection between 2016 and 2021 are reviewed. The reviewed papers are summarized in 4 figures and 10 tables that present detailed information about 29 pre-trained CNN models, 13 DR data sets, and standard performance metrics.Item Parkinson's Disease Monitoring from Gait Analysis via Foot-Worn Sensors(2018) Asuroglu, Tunc; Acici, Koray; Erdas, Cagatay Berke; Toprak, Munire Kilinc; Erdem, Hamit; Ogul, Hasan; https://orcid.org/0000-0002-3821-6419; https://orcid.org/0000-0001-7979-0276; AAC-7834-2020; HDM-9910-2022; AAJ-8674-2021Background: In Parkinson's disease (PD), neuronal loss in the substantia nigra ultimate in dopaminergic denervation of the stiratum is followed by disarraying of the movements' preciseness, automatism, and agility. Hence, the seminal sign of PD is a change in motor performance of affected individuals. As PD is a neurodegenerative disease, progression of disability in mobility is an inevitable consequence. Indeed, the major cause of morbidity and mortality among patients with PD is the motor changes restricting their functional independence. Therefore, monitoring the manifestations of the disease is crucial to detect any worsening of symptoms timely, in order to maintain and improve the quality of life of these patients. Aim: The changes in motion of patients with PD can be ascertained by the help of wearable sensors attached to the limbs of subjects. Then analysing the recorded data for variation of signals would make it possible to figure an individualized profile of the disease. Advancement of such tools would improve understanding of the disease evolution in the long term and simplify the detection of precipitous changes in gait on a daily basis in the short term. In both cases the apperception of such events would contribute to improve the clinical decision making process with reliable data. To this end, we offer here a computational solution for effective monitoring of PD patients from gait analysis via multiple foot-worn sensors. Methods: We introduce a supervised model that is fed by ground reaction force (GRF) signals acquired from these gait sensors. We offer a hybrid model, called Locally Weighted Random Forest (LWRF), for regression analysis over the numerical features extracted from input signals to predict the severity of PD symptoms in terms of Universal Parkinson Disease Rating Scale (UPDRS) and Hoehn and Yahr (H&Y) scale. From GRF signals sixteen time-domain features and seven frequency-domain features were extracted and used. Results and conclusion: An experimental analysis conducted on a real data acquired from PD patients and healthy controls has shown that the predictions are highly correlated with the clinical annotations. Proposed approach for severity detection has the best correlation coefficient (CC), mean absolute error (MAE) and root mean squared error (RMSE) values with 0.895, 4.462 and 7.382 respectively in terms of UPDRS. The regression results for H&Y Scale discerns that proposed model outperforms other models with CC, MAE and RMSE with values 0.960, 0.168 and 0.306 respectively. In classification setup, proposed approach achieves higher accuracy in comparison with other studies with accuracy and specificity of 99.0% and 99.5% respectively. Main novelty of this approach is the fact that an exact value of the symptom level can be inferred rather than a categorical result that defines the severity of motor disorders. (C) 2018 Nalecz Institute of Biocybernetics and Biomedical Engineering of the Polish Academy of Sciences. Published by Elsevier B.V. All rights reserved.Item Tuning of Output Scaling Factor in PI-Like Fuzzy Controllers for Power Converters Using PSO(2016) Erdem, Hamit; Altinoz, Okkes TolgaProportional-Integral (PI) like Fuzzy Logic Controllers (FLC) has been widely used for control of static power converters (SPC). The performance of these controllers is sensitive to controller rules, parameters of membership functions and input-output scaling factors. Among these parameters, scaling factor (SF) directly affects the controller performance in terms of transient response, steady state error and stability. Therefore, using an optimum SF value increases the performance of the FLC against using a constant value. Hence, in this paper optimizing the output scaling factor (OSF) of the PI-like fuzzy logic controller (PIFLC) by using Particle Swarm Optimization (PSO) algorithm is proposed. In order to optimize and analyze the effect of this parameter on the controller performance, first the output scaling factor of FLC is optimized with various PSO algorithms and one of these algorithms is selected for experimental test. Then optimized FLC is applied to a DC-DC Buck converter, and the performance of the controller is evaluated under nominal load and load disturbance. The controller design, the OSF optimization, and the controller performance analysis approaches are presented in detail.Item Turkish Speech Recognition Techniques and Applications of Recurrent Units (LSTM and GRU)(2021) Tombaloglu, Burak; Erdem, HamitA typical solution of Automatic Speech Recognition (ASR) problems is realized by feature extraction, feature classification, acoustic modeling and language modeling steps. In classification and modeling steps, Deep Learning Methods have become popular and give more successful recognition results than conventional methods. In this study, an application for solving ASR problem in Turkish Language has been developed. The data sets and studies related to Turkish Language ASR problem are examined. Language models in the ASR problems of agglutative language groups such as Turkish, Finnish and Hungarian are examined. Subword based model is chosen in order not to decrease recognition performance and prevent large vocabulary. The recogniton performance is increased by Deep Learning Methods called Long Short Term Memory (LSTM) Neural Networks and Gated Recurrent Unit (GRU) in the classification and acoustic modeling steps. The recognition performances of systems including LSTM and GRU are compared with the the previous studies using traditional methods and Deep Neural Networks. When the results were evaluated, it is seen that LSTM and GRU based Speech Recognizers performs better than the recognizers with previous methods. Final Word Error Rate (WER) values were obtained for LSTM and GRU as 10,65% and 11,25%, respectively. GRU based systems have similar performance when compared to LSTM based systems. However, it has been observed that the training periods are short. Computation times are 73.518 and 61.020 seconds respectively. The study gave detailed information about the applicability of the latest methods to Turkish ASR research and applications.Item Sparsity-driven weighted ensemble classifier(2018) Erdem, Hamit; Ozgur, Atilla; Nar, FatihIn this study, a novel sparsity-driven weighted ensemble classifier (SDWEC) that improves classification accuracy and minimizes the number of classifiers is proposed. Using pre-trained classifiers, an ensemble in which base classifiers votes according to assigned weights is formed. These assigned weights directly affect classifier accuracy. In the proposed method, ensemble weights finding problem is modeled as a cost function with the following terms: (a) a data fidelity term aiming to decrease misclassification rate, (b) a sparsity term aiming to decrease the number of classifiers, and (c) a non-negativity constraint on the weights of the classifiers. As the proposed cost function is non-convex thus hard to solve, convex relaxation techniques and novel approximations are employed to obtain a numerically efficient solution. Sparsity term of cost function allows trade-off between accuracy and testing time when needed. The efficiency of SDWEC was tested on 11 datasets and compared with the state-of-the art classifier ensemble methods. The results show that SDWEC provides better or similar accuracy levels using fewer classifiers and reduces testing time for ensemble.