Fatos Vural - Academia.edu (original) (raw)

Papers by Fatos Vural

Research Square (Research Square), Feb 22, 2024

Although Convolutional Neural Networks (CNN) outperform the classical models in a wide range of M... more Although Convolutional Neural Networks (CNN) outperform the classical models in a wide range of Machine Vision applications, their restricted interpretability and their lack of comprehensibility in reasoning, generates many problems such as security, reliability and safety. Consequently, there is a growing need on research to improve explainability and address their limitations. In this paper, we propose a concept-based method, called Concept-Aware Explainability (CAE) to provide a verbal explanation for the predictions of pre-trained CNN models. A new measure, called detection score mean, is introduced to quantify the relationship between the filters of the model and a set of pre-defined concepts. Based on the detection score mean values, we define sorted lists of Concept-Aware Filters (CAF) and Filter-Activating Concepts (FAC). These lists are used to generate explainability reports, where we can explain, analyze, and compare models in terms of the concepts embedded in the image. The proposed explainability method is compared to the state-of-the-art methods to explain Resnet18 and VGG16 models, pre-trained on ImageNet and Places365-Standard datasets. Two popular metrics, namely, number of unique detectors and number of detecting filters, are used to make a quantitative comparison. Superior performances are observed for the suggested CAE, when compared to Network Dissection (NetDis) [1] and Net2Vec [2] methods.

IEEE Access, 2021

Electroencephalography (EEG) serves as an effective diagnostic tool for mental disorders and neur... more Electroencephalography (EEG) serves as an effective diagnostic tool for mental disorders and neurological abnormalities. Enhanced analysis and classification of EEG signals can help improve performance in classifying the disorders and abnormalities. A new approach is examined here for enhancing EEG classification performance using a novel model of data representation that leverages knowledge of spatial layout of EEG sensors. An investigation of the performance of the proposed data representation model provides evidence of consistently higher classification accuracy of the proposed model compared with a model that ignores the sensor layout. The performance is assessed for models that represent the information content of the EEG signals in two different ways: a one-dimensional concatenation of the channels of the frequency bands and a proposed image-like two-dimensional representation of the EEG channel locations. The models are used in conjunction with different machine learning techniques. Performance of these models is examined on two tasks: social anxiety disorder classification, and emotion recognition using a dataset, DEAP, for emotion analysis using physiological signals. We hypothesize that the proposed two-dimensional model will significantly outperform the one-dimensional model and this is validated in our results as this model consistently yields 5-8% higher accuracy in all machine learning algorithms investigated. Among the algorithms investigated, Convolutional Neural Networks provide the best performance, far exceeding that of Support Vector Machine and k-Nearest Neighbors algorithms.

Computer Speech & Language, Sep 1, 2017

We propose multi-way, multilingual neural machine translation. The proposed approach enables a si... more We propose multi-way, multilingual neural machine translation. The proposed approach enables a single neural translation model to translate between multiple languages, with a number of parameters that grows only linearly with the number of languages. This is made possible by having a single attention mechanism that is shared across all language pairs. We train the proposed multi-way, multilingual model on ten language pairs from WMT'15 simultaneously and observe clear performance improvements over models trained on only one language pair. We empirically evaluate the proposed model on low-resource language translation tasks. In particular, we observe that the proposed multilingual model outperforms strong conventional statistical machine translation systems on TurkishÀEnglish and UzbekÀEnglish by incorporating the resources of other language pairs.

Turkish Journal of Electrical Engineering and Computer Sciences, Mar 1, 2005

A major design issue in content-based image retrieval system is the selection of the feature set.... more A major design issue in content-based image retrieval system is the selection of the feature set. This study attacks the problem of finding a discriminative feature for each class, which is optimal in some sense. Fuzzy ARTMAP architecture is used to find this discriminative feature set. For this purpose, initially, a large variety of features are extracted from the regions of the pre-segmented images. Then, the feature set of each object class is learned using the Fuzzy Art Map Architecture, by identifying the weights of each feature for each object class. In the querying phase, trained set of feature weights of fuzzy ARTMAP's are used to find the label of each object class. This task is achieved by combining the regions in the images and computing the maximum membership value for the compound regions, which correspond to a possible object class. The query object is matched to each segment group in a fuzzy database using the membership values of segment groups.

In this study, we aim to measure the information content of anatomic regions using the functional... more In this study, we aim to measure the information content of anatomic regions using the functional magnetic resonance images recorded during complex problem solving (CPS) task. We propose an information theoretic method for analyzing the activity in anatomic regions. We estimate two types of Shannon entropy, namely, static and dynamic entropy, and investigate the relationship between the CPS task phases and the entropy measures for the underlying brain activity. We propose a novel method to estimate static and dynamic brain networks using Kulback-Leibler divergence and investigate the validity of the estimated brain networks by modeling two main phases of complex problem solving process. The suggested computational network model is tested using Support Vector Machines. The network models can successfully discriminate the planning and execution phases of CPS with more than 90% accuracy .

In this paper, we introduce graph simplification capabilities of a new tool, CEREBRA, which is us... more In this paper, we introduce graph simplification capabilities of a new tool, CEREBRA, which is used to visualize the 3D network of human brain, extracted from the fMRI data. The nodes of the network are defined as the voxels with the attributes corresponding to the intensity values changing by time and the coordinates in three dimensional Euclidean space. The arc weights are estimated by modeling the relationships among the voxel activation records. We aim to help researchers to reveal the underlying brain state by examining the active regions of the brain and observe the interactions among them. Although the tool provides many features for displaying the fMRI data as a dynamical network, in this study, we have mainly focused on two main features. The first one is the unique graph simplification module that allows users to eliminate redundant edges according to some weighted similarity criterion. The second one is visualizing the output of the external algorithms for voxel selection, clustering or network representation of fMRI data. Thus, users are able to display, analyze and further process the output of their own algorithms.

The problem of detecting the presence of Social Anxiety Disorder (SAD) using Electroencephalograp... more The problem of detecting the presence of Social Anxiety Disorder (SAD) using Electroencephalography (EEG) for classification has seen limited study and is addressed with a new approach that seeks to exploit the knowledge of EEG sensor spatial configuration. Two classification models, one which ignores the configuration (model 1) and one that exploits it with different interpolation methods (model 2), are studied. Performance of these two models is examined for analyzing 34 EEG data channels each consisting of five frequency bands and further decomposed with a filter bank. The data are collected from 64 subjects consisting of healthy controls and patients with SAD. Validity of our hypothesis that model 2 will significantly outperform model 1 is borne out in the results, with accuracy 6-7% higher for model 2 for each machine learning algorithm we investigated. Convolutional Neural Networks (CNN) were found to provide much better performance than SVM and kNNs.

Signal, Image and Video Processing

Frontiers in Neuroinformatics

2022 IEEE 22nd International Conference on Bioinformatics and Bioengineering (BIBE)

2022 30th Signal Processing and Communications Applications Conference (SIU)

2020 25th International Conference on Pattern Recognition (ICPR)

Representing brain activities by networks is very crucial to understand various cognitive states.... more Representing brain activities by networks is very crucial to understand various cognitive states. This study proposes a novel method to estimate static and dynamic brain networks using Kulback-Leibler divergence. The suggested brain networks are based on the probability distributions of voxel intensity values measured by functional Magnetic Resonance Images (fMRI) recorded while the subjects perform a predefined cognitive task, called complex problem solving. We investigate the validity of the estimated brain networks by modeling and analyzing the different phases of complex problem solving process of human brain, namely planning and execution phases. The suggested computational network model is tested by a classification schema using Support Vector Machines. We observe that the network models can successfully discriminate the planning and execution phases of complex problem solving process with more than 90% accuracy, when the estimated dynamic networks, extracted from the fMRI data, are classified by Support Vector Machines.

Türkiye Bilişim Derneği (TBD), 2020

2019 9th International IEEE/EMBS Conference on Neural Engineering (NER), 2019

Lecture Notes in Computer Science, 2018

In this study, we propose a novel brain parcellation algorithm, called BrainParcel. BrainParcel d... more In this study, we propose a novel brain parcellation algorithm, called BrainParcel. BrainParcel defines a set of supervoxels by partitioning a voxel level brain graph into a number of subgraphs, which are assumed to represent “homogeneous” brain regions with respect to a predefined criteria. Aforementioned brain graph is constructed by a set of local meshes, called mesh networks. Then, the supervoxels are obtained using a graph partitioning algorithm. The supervoxels form partitions of brain as an alternative to anatomical regions (AAL). Compared to AAL, supervoxels gather functionally and spatially close voxels. This study shows that BrainParcel can achieve higher accuracies in cognitive state classification compared to AAL. It has a better representation power compared to similar brain segmentation methods, reported the literature.

2017 21st National Biomedical Engineering Meeting (BIYOMUT), 2017

Bu çalışma sıkıştırılmış manyetik rezonans görüntülerinin (MRG) ortak sözlükögrenimi kullanılarak... more Bu çalışma sıkıştırılmış manyetik rezonans görüntülerinin (MRG) ortak sözlükögrenimi kullanılarak geri-kazanımını incelemektedir. Genellikle sıkıştırılmış algılama içinönceden belirlenmiş sözlükler kullanılır. Burada, görüntü ve seyrelten dönüşümün yalnızca veri kullanılarak geri-kazanılması için dönüşümlü minimizasyon tabanlı bir algoritmaönerilmektedir. Onerilen yöntem aynı zamanda dahaönceden kullanılan bir ongörüsüz sıkıştırılmış algılama metodunun ortak geri-kazanım için genişletilmesi şeklinde de görülebilir [1]. Algoritma başarımınıölçmek için, algoritma yakınsama hızı ve görüntü kalitesi olarak hem ayrık sözlükögrenimi tabanlı metod [1] ile hem deönceden belirlenmiş sözlükleri kullanan bir MRG için ortak geri-kazanım algoritması ile kıyaslanmıştır [2].

2021 29th Signal Processing and Communications Applications Conference (SIU), 2021

Machine Vision Applications in Industrial Inspection V, 1997

This paper deals with a class of textures which can be represented by Markov Random Fields (MRF) ... more This paper deals with a class of textures which can be represented by Markov Random Fields (MRF) model. It is well known that by changing the MRF parameters, extremely wide group of textures can be generated. However, it is not easy to model and classify a textured image, since there is no clear-cut mathematical definition of texture. Although, many classification methods exist in the literature, the success of the results heavily depends on the data type. Thus, appropriate measures which give visually meaningful representation of texture are highly desirable. In this study a new set of texture measures, namely, Mean Clique Length (MCL) and Clique Standard Deviation (CSD) is introduced. These measures are defined employing new concepts which agrees with the human visual system. The simulation experiments are performed on binary MRF texture alphabet to quantify the data by the MCL and CSD measures. Experimental results indicate that the introduced measures identify the visually similar textures much better than the mathematical distance measures.

Proceedings 2001 International Conference on Image Processing (Cat. No.01CH37205)

In this paper, a new texture descriptor, namely, Statistical Analysis of Structural Information (... more In this paper, a new texture descriptor, namely, Statistical Analysis of Structural Information (SASI) is introduced as a representation of texture. SASI is based on statistics of clique autocorrelation functions calculated over a set of directional moving windows. SASI defines a set of windows to extract and measure various structural properties of texture by using a spatial multiresolution method. Although it works in spatial domain, it measures the spectral information of a given texture. Experimental results, performed on digitized Brodatz Album, indicate that SASI is very successful in identifying the "similar" textures.

Speech Communication, 1988

ABSTRACT