Elektrik Elektronik Mühendisliği Bölümü Koleksiyonu

Bu koleksiyon için kalıcı URI

https://hdl.handle.net/20.500.12491/144

Listeleniyor 1 - 20 / 271

Residual block fully connected DCNN with categorical generalized focal dice loss and its application to Alzheimer's disease severity detection
(PeerJ Inc, 2023) Alhudhaif, Adi; Polat, Kemal
Background. Alzheimer's disease (AD) is a disease that manifests itself with a deteriora-tion in all mental activities, daily activities, and behaviors, especially memory, due to the constantly increasing damage to some parts of the brain as people age. Detecting AD at an early stage is a significant challenge. Various diagnostic devices are used to diagnose AD. Magnetic Resonance Images (MRI) devices are widely used to analyze and classify the stages of AD. However, the time-consuming process of recording the affected areas of the brain in the images obtained from these devices is another challenge. Therefore, conventional techniques cannot detect the early stage of AD.Methods. In this study, we proposed a deep learning model supported by a fusion loss model that includes fully connected layers and residual blocks to solve the above -mentioned challenges. The proposed model has been trained and tested on the publicly available T1-weighted MRI-based KAGGLE dataset. Data augmentation techniques were used after various preliminary operations were applied to the data set.Results. The proposed model effectively classified four AD classes in the KAGGLE dataset. The proposed model reached the test accuracy of 0.973 in binary classification and 0.982 in multi-class classification thanks to experimental studies and provided a superior classification performance than other studies in the literature. The proposed method can be used online to detect AD and has the feature of a system that will help doctors in the decision-making process.
A novel adaptive time delay identification technique
(Elsevier Science Inc, 2023) Bayrak, Alper; Tatlıcıoğlu, Enver
In this work, a novel online adaptive time delay identification method is proposed for a class of signal processing and communication applications, where the received signal is a combination of the transmitted signal and its delayed forms with the delay values being uncertain and required to be estimated. The design relies on a filtered form of a prediction error-like term which is then used in the design of the novel nonlinear adaptive update law. The stability of the identification algorithm is then investigated via novel Lyapunov based tools and the time delay identification is proven to be globally uniformly ultimately bounded. Several numerical simulations are conducted to evaluate the performance of the proposed identifier where constant, slowly varying and suddenly changing delays are successfully identified even in the presence of additive noise.& COPY; 2023 Published by Elsevier Ltd on behalf of ISA.
Investigation of energy generation potential of solar panels placed as shutters for windows in residential
(Taylor & Francis Inc, 2023) Fidan, Murat
This study investigated the annual potential to generate electricity from 260 W solar panels placed on windows as a shutter of a sample flat in a mass housing in Bolu, Turkiye, which has adverse conditions such as shadowing and limited sunshine duration. Results showed that 4.36% of the annual energy consumed by the sample flat from the network can be produced by solar panels placed on the windows as shutter. It is predicted that the residential windows in the 500 flats mass housing where the sample flat is located have the potential to generate at least 53.73 MWh of electrical energy per year. This generated energy can prevent the emission of 14,010,632 tons of CO2 annually. It has been shown that the electrical energy produced from the residential windows can be used with 94.55% efficiency with the DC installation proposed within the scope of the study. It is shown that spread lighting provided by LED strips may be an alternative for Light Fidelity communication. This is the first study for Bolu in which measurements were recorded for one year.
Improving energy efficiency in climatic test chambers with deep learning and absolute humidity methods
(IEEE, 2023) Bekiroğlu, Erdal; Karaca, Hakan
Climatic test chambers are devices used to simulate environmental conditions for the testing and verification of products in various industries. However, these chambers can consume significant amounts of energy, resulting in high operating costs and environmental impacts. Therefore, the need to optimize the energy efficiency of climatic test chambers while maintaining their performance is becoming increasingly important. In this paper, we will discuss the control method for humidity testing by calculating the use of the LSTM algorithm instead of the classical control method PID to control climatic test chambers to improve energy efficiency and the control method based on absolute humidity instead of relative humidity. In particular, we harness the power of artificial neural networks to reduce energy consumption and improve control of climatic test chambers based on various input parameters such as temperature, humidity, and test duration. By changing the control methods, we aim to increase efficiency and make it more suitable and efficient for smart grid systems.
Design and magnetic analysis of a grounding transformer compatible with wind power system
(IEEE, 2023) Demiral, Elif; Bekiroğlu, Erdal; Yazar, Muhammed Duran
The aim of this study is to design and magnetic analysis of a grounding transformer in order to prevent material damage caused by phase-ground faults that may occur in wind energy systems. A zig-zag transformer is designed with a new winding design technique, different from the traditional zig-zag transformer design. Due to the lack of analytical formulas, finite element analysis (FEA) is used to calculate leakage reactance of designed transformer. The designed zig-zag transformer is analyzed in the ANSYS Maxwell environment. The proposed grounding transformer for wind power systems is investigated in terms of 3-phase 3D leakage reactance, 1-phase 3D leakage reactance, core losses of 3D, core losses of 2D, and flux density. The obtained results showed that the designed zig-zag transformer is compatible with wind power systems.
Stable measurement system for platinum resistance temperature detector
(Maik Nauka/Interperiodica/Springer, 2023) Altınkaya, Serdar; Bayrak, Alper; Daldal, Nihat; Özdil, Osman Eren
-Stable and precise temperature measurement is crucial for thermodynamic applications. The stability and precision of the measurement depend on the sensor type and measurement method. In industrial applications, platinum resistance temperature detectors (RTD) are widely preferred. In this study, stable and precise temperature measurement methods by using the Platinum RTD PT1000 sensor are investigated. Measurement methods are considered as measurement circuits and digital filter applications, separately. Experimental studies were evaluated on a household-type oven and comparative results are presented.
Tinba: Incremental partitioning for efficient trajectory analytics
(Elsevier Sci LTD, 2023) Tian, Ruijie; Zhang, Weishi; Wang, Fei; Polat, Kemal; Alenezi, Fayadh
Applications with mobile and sensing devices have already become ubiquitous. In most of these applications, trajectory data is continuously growing to huge volumes. Existing systems for big trajectory data organize trajectories at distributed block storage systems. Systems like DITA that use block storage (e.g., 128 MB each) are more efficient for analytical queries, but they cannot incrementally maintain the partitioned data and do not support delete operations, resulting in difficulties in trajectory analytics. In this paper, we propose an incremental trajectory partitioning framework Tinba that enables distributed block storage systems to efficiently maintain optimized partitions under incremental updates of trajectories. We employ a data flushing technique to bulk ingest trajectory data for random writing in distributed file system (DFS). We recast the incremental partitioning problem as an optimal partitioning problem and prove its NP-hardness. A cost- benefit model is proposed to address the optimal partitioning problem. Moreover, Tinba supports most of the existing similarity measures to quantify the similarity between trajectories. A heuristic is developed to instantiate the Tinba framework. Comprehensive experiments on real-world and synthetic datasets demonstrate the advancements in ingestion performance and partition quality, as opposed to other trajectory partition methods.
An ensemble learning approach for resampling forgery detection using Markov process
(Elsevier, 2023) Mehta, Rachna; Kumar, Karan; Alhudhaif, Adi; Alenezi, Fayadh; Polat, Kemal
Resampling is an extremely well-known technique performed for Image forgery detection. It includes the changes in the content of a picture in terms of rotation, stretching/zooming, and shrinking, to frame a forged picture that is a localized forgery in comparison to the original picture. With the wrong intention, resampling forgery has been increased day by day, and its negative impact has been increased in criminology, law enforcement, forensics, research etc. Accordingly, the interest in the algorithm of image resampling forgery detection is significantly developed in image forensics. In this paper, a novel image resampling forgery detection technique has been proposed. In the proposed technique, two types of Markov feature with spatial and Discrete Cosine Transform domains have been extracted to recognize the resampling operation. The spatial domain gives the information for the distribution of the pixels and DCT gives the edge information. Further, these Markov features are consolidated. Due to high dimensionality hard thresholding technique is used for reducing the dimensionality. Then, these Markov features are applied to the set of models of different classifiers. With the utilization of classifiers, weighted majority voting values have been calculated during the ensemble classification. Unlike the other techniques, these weighted voting boundaries have been consequently balanced during the training process until the best accuracy has been obtained. However, it is very difficult to get best accuracy so for getting best accuracy this research needs to do lots of iterations and trained the dataset. For the comparative study very few research has been found for this resampling forgery technique with different interpolation techniques and classifier. Still, comparison has been done with some latest research work. The comparative analysis shows that the proposed ensemble learning-based algorithm provides the best outcomes with the accuracy of 99.12% for bicubic, 98.89% for bilinear, and 98.23% for lanczos3 kernel with considerably less complexity and high speed in comparison to prior techniques which are using single support vector machine for classification. Moreover, the proposed algorithm also detects a very low probability of error of 0.44% and detects the type of interpolation kernel, size of the forgery, and the type of resampling, whether it is up sampling and down sampling, using Graphical User Interface which has not been detected previously with multiple forgery detection.& COPY; 2023 Elsevier B.V. All rights reserved.
Annealing-induced modifications on structural, surface chemical bonding, and electrical characteristics of p-NiO/n-TiO2 heterostructure
(Springer, 2023) Kaya, Şenol; Soykan, Uğur; Sunkar, Mustafa; Karaboğa, Seda; Doğan, Muhsin Uğur; Terzioğlu, Rıfkı; Yıldırım, Gürcan; Terzioğlu, Cabir
The influences of annealing temperatures on the electrical characteristics of a p- NiO/n-TiO2 heterojunction diode were thoroughly investigated, taking into account changes in microstructure, morphology, and surface chemistry of the p-NiO/n-TiO2 films, which were deposited on an insulating SiO2/ Si layer. During different annealing processes, considerable stress variations were observed in the p-NiO/n-TiO2 films due to the crystalline evolution of p-NiO and n-TiO2. Notably, the crystallization of the TiO2 layer, which serves as the intermediary between the back contact materials and NiO, led to the evident formation of grain structures. As the annealing temperature increased, the surface roughness also grew from 5.4 to 8.7 nm. At an annealing temperature of 500 degrees C, the formation of a parasitic NiTiOx phase was observed, particularly at the interface between NiO and TiO2. Conversely, the study also revealed that annealing temperature played a significant role in the rectifying behavior, barrier potential, and ideality factor of the diode. Among the various annealing processes, the most favorable results were achieved after annealing at 400 degrees C. At this temperature, the diode demonstrated the lowest ideality factor of 1.89, accompanied by superior rectifying behavior and a barrier potential of 0.70 eV. The findings clearly indicate that any alterations in the surface chemistry and microstructure of the film directly impact the diode's characteristics. Thus, optimizing the annealing temperature becomes crucial for enhancing the performance of the p-NiO/n-TiO2 heterojunction diode.
Technical feasibility of offshore wind power plant for Gokceada region in Turkiye
(IEEE, 2023) Akyol, Burak; Yazar, Muhammed Duran; Bekiroğlu, Erdal
The aim of this study is technical feasibility of the offshore wind power plant (OWPP) installation in Gokceada/Turkiye. First of all, the site selection process for OWPP has been executed. Then, the suitable wind turbine model for the wind speed profile of the region is determined. The layout of the turbines is made according to the main wind direction of the region. The designed system in the Homer PRO environment has been analyzed under actual wind speed. The ratio of the power produced by the system to meet the actual electricity demand of Canakkale and Turkish power system has been analyzed. In addition, the designed system is also analyzed under actual wind speed in the WindSim environment. Annual power generation, capacity factor, and full power generation time results are obtained. The applicability of the proposed system has been validated with the help of simulation environments. The obtained results showed that the OWPP in Gokceada/Turkiye is prevailing, reliable, and efficient.
Design and double-stage optimization of synchronous reluctance motor for electric vehicles
(Taylor & Francis Inc, 2023) Bekiroğlu, Erdal; Esmer, Sadullah
In this study, a high-power synchronous reluctance motor (SynRM) was designed for the traction motor of electric vehicle (EV) and its double-stage optimization was performed. Genetic algorithm and sensitivity analysis methods were used to obtain the best design parameters. Double-stage optimization was carried out to minimize the torque ripple and obtain the targeted torque, speed, and power values of the SynRM. In the first stage, the genetic algorithm method was used to improve the design parameters of the stator and rotor. With the improved design parameters, it was observed that the torque ripple decreased. In the second stage, the sensitivity analysis method was used. In this method, the effect of changing the skew angle of the stator on the torque ripple was investigated. The performance of the designed motor was examined in the optimization process. It was observed that the targeted torque, power, speed, efficiency, and torque ripple minimization values are successfully achieved with the best stator and rotor parameters. The results showed that SynRM produces high torque and high power with high efficiency and low torque ripple over wide speed range. It is quite proper to use the designed SynRM as a traction motor of new generation electric vehicles.
Near real-time load forecasting of power system using fuzzy time series, artificial neural networks, and wavelet transform models
(Taylor & Francis Inc., 2024) Khatoon, Shahida; Ibraheem, Mohammad; Shahid, Mohammad; Çelik, Emre; Bekiroğlu, Erdal; Sharma, Gulshan
Due to the increasing usage of electrical power, the size of electrical power system has increased manifold over the years. There is no inventory or buffer from generation to customer; therefore, to provide a reliable and quality electrical energy whenever demanded, power utility engineers require an adequate, efficient, and precise load forecast to meet continuously varying load demands. This article presents the design and analysis of demand forecasting over shorter interval for power system. The fuzzy time series (FTS), artificial neural network (ANN), and wavelet transform (WT) based forecasting is presented and analyzed in this article. The real-time data from Indian utility is collected for forecasting the demand and to check the effectiveness of FTS, ANN, and WT. The various error definitions are used to calculate the accuracy of the proposed techniques, and the application results verify the superiority of WT and ANN over FTS by showing reduced error value with greater accuracy. Additionally, it is watched that wavelet db3, level 3 is discovered to be the most accurate Daubechies wavelet-oriented technique for predicting the demand in comparison to other dbs, and it highly aligns in reducing the error between actual and predicted demand.
Accurate 3D contrast-free myocardial infarction delineation using a 4D dual-stream spatiotemporal feature learning framework
(Elsevier, 2023) Liu, Jinhao; Zhu, Xinglai; Xu, Chenchu; Xu, Lei; Polat, Kemal; Alenezi, Fayadh
Accurate 3D contrast-free myocardial infarction (MI) delineation has the potential to eliminate the need for toxic injections, thereby significantly advances diagnosis and treatment of MI. In this study, we propose a 4D dual-stream spatiotemporal feature learning framework (4D-DSS) that enables learning of 4D (3D + T) representation of the heart to accurately map the 3D MI regions, thereby directly delineating of 3D MI without contrast agent. This framework creatively introduces a dual-stream 3D spatiotemporal point cloud architecture enables to learn the myocardial 4D representation in both local and global aspects, and improve the comprehension and precision of the representation. Specifically, the framework utilizes the local spatiotemporal variation of individual point clouds to characterize minute distortions in myocardial regions and the global spatiotemporal variation of point cloud sequences to represent the overall myocardial motion between frames, thereby enables comprehensive learning of 3D myocardial motion and leverages these features to classify myocardial tissue into MI regions and normal regions. 4D-DSS significantly improved performance (with a precision increase of at least 4%) compared to four advanced methods. The results support the impact of our 4D-DSS framework on the development and implementation of 3D contrast-free myocardial infarction region delineation technology.& COPY; 2023 Elsevier B.V. All rights reserved.
Brain tumor classification using the modified ResNet50 model based on transfer learning
(Elsevier Sci LTD, 2023) Sharma, Arpit Kumar; Nandal, Amita; Dhaka, Arvind; Zhou, Liang; Alenezi, Fayadh; Polat, Kemal
Brain tumour classification is essential for determining the type and grade and deciding on therapy appropriately. Several diagnostic methods are used in the therapeutic therapy to identify brain tumours. MRI, on the other hand, offers superior picture clarity, which is why specialists depend on it. Furthermore, detecting cancer through the manual division of brain tumours is a time-consuming, exhausting, and difficult job. The handdesigned outlines for planned brain tumour growth methods are present in the majority of the instances. Segmentation is a highly reliable and precise method for assessing therapy prognosis, planning, and outcomes. Artificial Intelligence (AI), Machine Learning (ML), and Deep Learning (DL) advancements have enabled us to investigate the illness with high precision in a short period of time. Such technologies have produced some remarkable results, particularly in the last twenty years. Such breakthroughs provide doctors with the ability to evaluate the human anatomy using high-resolution sections. The most recent approaches can improve diagnostic precision when examining patients using non-invasive means. This work introduces a brain tumour detection method. The model grows using ResNet50, feature extraction, and augmentation. CNN's pre-trained datasets are used to fine-tune transfer learning. The proposed design utilised elements of the ResNet50 model, removing the final layer and adding four additional layers to meet work conditions. This study uses the improved ResNet50 model to present a novel deep-learning approach based on a transfer learning technique for evaluating brain cancer categorisation accuracy. Performance metrics were used to evaluate the effectiveness of the proposed model, and the results were compared to those obtained using state-of-the-art methods.
Secrecy outage probability and limiting threshold criteria in an optimal SNR regime by maximizing desired transfer rate
(Elsevier, 2023) Polat, Kemal
In this paper, the optimum secrecy probability has been calculated using a reduction function that optimizes the transfer rate of the user's signal for a given average broadcast power while mini-mizing the transfer rate of the eavesdropper's signal to ensure safe transmission. In this work, the signal-to-noise ratio (SNR) of the eavesdropper signal and the received signal are accurately estimated using matrix theory. These approximated SNRs have been studied in relation to the outage probability. By making three distinct assumptions about channel correlation, it has been possible to determine the precise secrecy outage probability. In correlated multi-antenna transmit and receive channels, this research investigates the secrecy performance of three popular di-versity combining strategies-maximum ratio combining (MRC), selection combining (SC), and equal gain combining (EGC). In this research, we consider a Rayleigh fading channel with overhead communication between the transmitter and recipient. In this research, we also propose a limiting bounds condition to improve the limits on the outage probability for the Rayleigh fading environment using the three diversity combining methods. Our restricting distribution is threshold-based in order to generate precise limits on the probability of secrecy outage. An application case has been presented which discusses the impact of the proposed model for device to device communication application scenario. Our findings demonstrate that channel correlation substantially influences the probability of a secrecy outage.
Peak shaving control of EV charge station with a flywheel energy storage system in micro grid
(IEEE, 2023) Bekiroğlu, Erdal; Esmer, Sadullah
In this study, peak shaving control is applied for load balancing using micro grid and flywheel energy storage system (FESS). The proposed method is applied to an EV charge station with an unbalanced load curve. FESS is charged during the low energy demand period and discharged during the high energy demand period. The FESS is determined by considering the required energy capacity, input power, and output power. When the Peak shaving control has been performed using the specified FESS, it has been noticed that an unbalanced load between 7 and 230 kW could be successfully balanced. It has been observed that the balanced load fluctuates between 13 and 123 kW. Thus, a more balanced load demand for the EV charge station has been generated from the electricity grid.
Experimental and simulation study of pulse current generator
(Taylor & Francis Inc, 2024) Tunç, Emre; Fidan, Murat
8/20 mu s lightning pulse current generators (LPCG) used in testing surge arresters require a laborious production process that needs a high budget. An easily repairable, portable, and affordable test system with a maximum energy of 2.9 kJ and a voltage level of 24 kV that can alter its parameter values was described in this work. Simulation and experimental studies were carried out to investigate the effects of the resistor and coil elements in the Pulse Current Generator (PCG) circuit on the amplitude, front, and tail time of the pulse current. Most components of the experimental system were created using 3D printing technology with polylactic acid (PLA) material. The production costs were significantly reduced, according to equivalent pulse generators. Tests at different pulse currents and voltage levels were performed with the experimental system. Power capacitors may often fail due to overvoltage, switching, and partial discharge reasons. The high voltage (HV) power capacitors used in the pulse generator introduced in the study were obtained with the help of low voltage capacitor stacks. As a result, a less expensive, high-capacity capacitor system that is easily and quickly repairable in case of failure was achieved.
MPPT control of grid connected DFIG at variable wind speed
(MDPI, 2022) Bekiroğlu, Erdal; Yazar, Muhammed Duran
In this study, maximum power point tracking (MPPT) control of a grid-connected doubly fed induction generated (DFIG)-based wind energy conversion system (WECS) at variable wind speed was designed and analyzed. The real wind speed data of the Edremit/Balikesir region in Turkey was used as the wind speed profile. A N90/2.5 MW wind turbine model of Nordex Company was used in the study. Firstly, a conventional PI controller was applied to both rotor and grid side converters. The rotor-side converter (RSC) controls the power generated from the DFIG, whereas the grid-side converter (GSC) controls the DC bus voltage. An MPPT controller was applied to the RSC to generate reference torque at instant variable wind speeds. Thus, the system's response time, electromagnetic torque, generated power, and grid-side currents parameters were improved. In the MPPT controller, the reference torque value is produced by using the angular velocity and reference angular velocity values of the DFIG. The proposed system was modeled and simulated in Matlab/Simulink. Generated power, DC bus voltage, response time, electromagnetic torque, and grid side currents results were obtained. The results of the conventional PI controller and the results of the PI controller with MPPT were compared. The results of the proposed control were also compared with the related studies. The results showed that the proposed system is reliable, applicable, and valid for the grid-connected DFIG at variable wind speeds.
Novel dual-channel long short-term memory compressed capsule networks for emotion recognition
(Pergamon-Elsevier Science Ltd, 2022) Shahin, Ismail; Hindawi, Noor; Nassif, Ali Bou; Alhudhaif, Adi; Polat, Kemal
Recent analysis on speech emotion recognition (SER) has made considerable advances with the use of MFCC's spectrogram features and the implementation of neural network approaches such as convolutional neural networks (CNNs). The fundamental issue of CNNs is that the spatial information is not recorded in spectrograms. Capsule networks (CapsNet) have gained gratitude as alternatives to CNNs with their larger capacities for hierarchical representation. However, the concealed issue of CapsNet is the compression method that is employed in CNNs cannot be directly utilized in CapsNet. To address these issues, this research introduces a text-independent and speaker-independent SER novel architecture, where a dual-channel long short-term memory compressed-CapsNet (DC-LSTM COMP-CapsNet) algorithm is proposed based on the structural features of CapsNet. Our proposed novel classifier can ensure the energy efficiency of the model and adequate compression method in speech emotion recognition, which is not delivered through the original structure of a CapsNet. Moreover, the grid search (GS) approach is used to attain optimal solutions. Results witnessed an improved performance and reduction in the training and testing running time. The speech datasets used to evaluate our algorithm are: Arabic Emirati-accented corpus, English speech under simulated and actual stress (SUSAS) corpus, English Ryerson audio-visual database of emotional speech and song (RAVDESS) corpus, and crowd-sourced emotional multimodal actors dataset (CREMA-D). This work reveals that the optimum feature extraction method compared to other known methods is MFCCs delta-delta. Using the four datasets and the MFCCs delta-delta, DC-LSTM COMP-CapsNet surpasses all the state-of-the-art systems, classical classifiers, CNN, and the original CapsNet. Using the Arabic Emirati-accented corpus, our results demonstrate that the proposed work yields average emotion recognition accuracy of 89.3% compared to 84.7%, 82.2%, 69.8%, 69.2%, 53.8%, 42.6%, and 31.9% based on CapsNet, CNN, support vector machine (SVM), multi-layer perceptron (MLP), k-nearest neighbor (KNN), radial basis function (RBF), and naive Bayes (NB), respectively.
Emotional speaker identification using a novel capsule nets model
(Pergamon-Elsevier Science Ltd, 2022) Nassif, Ali Bou; Shahin, Ismail; Elnagar, Ashraf; Velayudhan, Divya; Alhudhaif, Adi; Polat, Kemal
Speaker recognition systems are widely used in various applications to identify a person by their voice; however, the high degree of variability in speech signals makes this a challenging task. Dealing with emotional variations is very difficult because emotions alter the voice characteristics of a person; thus, the acoustic features differ from those used to train models in a neutral environment. Therefore, speaker recognition models trained on neutral speech fail to correctly identify speakers under emotional stress. Although considerable advancements in speaker identification have been made using convolutional neural networks (CNN), CNNs cannot exploit the spatial association between low-level features. Inspired by the recent introduction of capsule networks (CapsNets), which are based on deep learning to overcome the inadequacy of CNNs in preserving the pose relationship between low-level features with their pooling technique, this study investigates the performance of using CapsNets in identifying speakers from emotional speech recordings. A CapsNet-based speaker identification model is proposed and evaluated using three distinct speech databases, i.e., the Emirati Speech Database, SUSAS Dataset, and RAVDESS (open-access). The proposed model is also compared to baseline systems. Experimental results demonstrate that the novel proposed CapsNet model trains faster and provides better results over current stateof-the-art schemes. The effect of the routing algorithm on speaker identification performance was also studied by varying the number of iterations, both with and without a decoder network.

Güncel Gönderiler