A5103272723- Face and Expression Recognition
- Speech and Audio Processing
- Music and Audio Processing
- Topic Modeling
- Sparse and Compressive Sensing Techniques
- Natural Language Processing Techniques
- Machine Learning and ELM
- Blind Source Separation Techniques
- Advanced Battery Technologies Research
- Neural Networks and Applications
- Spectroscopy and Chemometric Analyses
- Text and Document Classification Technologies
- Wind Turbine Control Systems
- Electric Vehicles and Infrastructure
- Power Systems Fault Detection
- Deception detection and forensic psychology
- Islanding Detection in Power Systems
- Radiomics and Machine Learning in Medical Imaging
- Digital and Cyber Forensics
- Web Data Mining and Analysis
- Energy and Environment Impacts
- COVID-19 diagnosis using AI
- Speech Recognition and Synthesis
- Photovoltaic Systems and Sustainability
- Power System Reliability and Maintenance
Necmettin Erbakan University
2020-2024
Ankara Sosyal Bilimler Üniversitesi
2022-2023
Sabancı Üniversitesi
2010-2022
In this paper, a novel approach for single channel source separation (SCSS) using deep neural network (DNN) architecture is introduced. Unlike previous studies in which DNN and other classifiers were used classifying time-frequency bins to obtain hard masks each source, we use the classify estimated spectra check their validity during separation. training stage, data signals are train DNN. trained utilized aid estimation of mixed signal. Single problem formulated as an energy minimization...
Electric vehicles (EVs), which are environmentally friendly, have been used to minimize the global warming caused by fossil fuels in and increasing fuel prices due decrease resources. Considering that energy EVs is obtained from resources, it also important store use efficiently EVs. In this context, recovery a regenerative braking system plays an role EV efficiency. This paper presents fuzzy logic-based hybrid storage technique consisting of supercapacitor (SC) battery for efficient safe...
Hearings of witnesses and defendants play a crucial role when reaching court trial decisions. Given the high-stakes nature outcomes, developing computational models that assist decision-making process is an important research venue. In this article, we address identification deception in real-life data. We use dataset consisting videos collected from public trials. explore verbal non-verbal modalities to build multimodal detection system aims discriminate between truthful deceptive...
Hate speech detection is a crucial task, especially on social media, where harmful content can spread quickly. Implementing machine learning models to automatically identify and address hate essential for mitigating its impact preventing proliferation. The first step in developing an effective model acquire high-quality dataset training. Labeled data foundational most natural language processing tasks, but categorizing difficult due the diverse often subjective nature of speech, which lead...
For classifier ensembles, an effective combination method is to combine the outputs of each using a linearly weighted rule. There are multiple ways and it beneficial analyze them as whole. We present unifying framework for linear types in this paper. This unification enables same learning algorithms different combiners. various train weights regularized empirical loss minimization. propose hinge better performance compared conventional least-squares loss. effects weight training by running...
Estimation of the wind speed makes a very important contribution to seamless integration power plants into grid. In this way, maximum amount electricity can be generated by estimating energy that from energy. The measurements in region, where plant is established, made before installation (WPP), takes between 6 and 18 months. study, it was investigated what could done make foresight estimation about future for selected region. order accurately determine speed, tried estimated using...
High-quality word representations have been very successful in recent years at improving performance across a variety of NLP tasks. These are the mappings each vocabulary to real vector Euclidean space. Besides high on specific tasks, learned shown perform well establishing linear relationships among words. The recently introduced skip-gram model improved unsupervised learning embeddings that contains rich syntactic and semantic relations both terms accuracy speed. Word used frequently...
Word embeddings are successfully employed in various Natural Language Processing tasks, but training them requires large amount of text, which is scarce for Turkish. In this work, we collected amounts articles from two news websites and tags within web pages used as labels. Obtained corpora tested with document classification models. Embedding based models performed better than the traditional TF-IDF features. A neural network that simultaneously learns word best.
In recent years, the energy crisis has become more and serious. Li-ion batteries are used in grids because of their benefits such as contributing to intermittent generation renewable sources stabilizing grid. addition, li-ion widely electric vehicles due long cycle life high density. battery state charge (SoC) is an important indicator for safety. Therefore, SoC estimation important. Today, there different methods determine many applications. The traditional method, ampere-hour integration...
In Diesel Engines, the heat energy obtained from glow plug increases engine's ability to start up in cold climatic conditions and significantly reduces emissions of harmful gases leaving exhaust. conditions, before off diesel vehicles it is necessary wait for about 10 s cylinder block heating. This period negatively affects driving comfort. this study, mathematical results processes optimize time required reach temperature have been experimentally proven. A test apparatus was developed...
Global warming is seen as one of the most important problems that trigger climate change in world.The leading cause global high amount greenhouse gases released into atmosphere.Countries are creating policies to encourage use renewable energy sources control gas emissions.Many universities Türkiye have enough campus areas generate electricity from sources.The main purpose this article analyze feasibility developing a solar power plant at Necmettin Erbakan University.This proposes 1MW...
"Tandem approach" is a method used in speech recognition to increase performance by using classifier posterior probabilities as observations hidden Markov model. In this work we study the effect of multiple visual tandem features improve audio-visual accuracy. addition, investigate methods combine outputs several audio and classifiers with fusion system generate learned weights. Experiments show that both approaches help respect regular especially noisy environments.
Classifier combination is an important research area since they have a significant contribution to the accuracy. Even though simple fixed rules result in satisfactory performances, supervised learning surely has higher probability of better Among methods, linear combiner one well-known methods. In this work, different types combiners are examined unifying framework and regularized least squares (LS) method investigated for combiner. Experiments conducted on three databases results examined....
Detecting COVID-19 in computed tomography (CT) or radiography images has been proposed as a supplement to the definitive RT-PCR test. We present deep learning ensemble for detecting infection, combining slice-based (2D) and volume-based (3D) approaches. The 2D system detects infection on each CT slice independently, them obtain patient-level decision via different methods (averaging long-short term memory networks). 3D takes whole volume arrive one step. A new high resolution chest scan...
Classifier combination has been an important research area because of their contribution to the accuracy and robustness. Supervised linear combiner types are shown be strong combiners; but nonlinear not well investigated. In this work, we show a method obtain non-linear versions simple types. Experiments conducted on four different databases results examined. It is observed that can better accuracies with combinations for certain