MOMete Ozay

Mete Ozay

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Mete Ozay has a total of 9 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Gender classification using mesh networks on multiresolution multitask fMRI data

    by Itir Onal Ertugrul, Mete Ozay, Fatos T. Yarman Vural

    Brain connectivity networks have been shown to represent gender differences under a number of cognitive tasks. Recently, it has been conjectured that fMRI signals decomposed into different resolutions embed different types of cognitive information. In this paper, we combine multiresolution analysis and connectivity networks to study gender differences under a variety of cognitive tasks, and propose a machine learning framework to discriminate individuals according to their gender. For this purpose, we estimate a set of brain networks, formed at different resolutions while the subjects perform different cognitive tasks. First, we decompose fMRI signals recorded under a sequence of cognitive stimuli into its frequency subbands using Discrete Wavelet Transform (DWT). Next, we represent the fMRI signals by mesh networks formed among the anatomic regions for each task experiment at each subband. The mesh networks are constructed by ensembling a set of local meshes, each of which represents the relationship of an anatomical region as a weighted linear combination of its neighbors. Then, we estimate the edge weights of each mesh by ridge regression. The proposed approach yields 2CL functional mesh networks for each subject, where C is the number of cognitive tasks and L is the number of subband signals obtained after wavelet decomposition. This approach enables one to classify gender under different cognitive tasks and different frequency subbands. The final step of the suggested framework is to fuse the complementary information of the mesh networks for each subject to discriminate the gender. We fuse the information embedded in mesh networks formed for different tasks and resolutions under a three-level fuzzy stacked generalization (FSG) architecture. In this architecture, different layers are responsible for fusion of diverse information obtained from different cognitive tasks and resolutions. In the experimental analyses, we use Human Connectome Project task fMRI dataset. Results reflect that fusing the mesh network representations computed at multiple resolutions for multiple tasks provides the best gender classification accuracy compared to the single subband task mesh networks or fusion of representations obtained using only multitask or only multiresolution data. Besides, mesh edge weights slightly outperform pairwise correlations between regions, and significantly outperform raw fMRI signals. In addition, we analyze the gender discriminative power of mesh edge weights for different tasks and resolutions.

  • Open Access

    Revisiting Single Image Depth Estimation: Toward Higher Resolution Maps with Accurate Object Boundaries

    by Junjie Hu, Mete Ozay, Yan Zhang, Takayuki Okatani

    This paper considers the problem of single image depth estimation. The employment of convolutional neural networks (CNNs) has recently brought about significant advancements in the research of this problem. However, most existing methods suffer from loss of spatial resolution in the estimated depth maps; a typical symptom is distorted and blurry reconstruction of object boundaries. In this paper, toward more accurate estimation with a focus on depth maps with higher spatial resolution, we propose two improvements to existing approaches. One is about the strategy of fusing features extracted at different scales, for which we propose an improved network architecture consisting of four modules: an encoder, decoder, multi-scale feature fusion module, and refinement module. The other is about loss functions for measuring inference errors used in training. We show that three loss terms, which measure errors in depth, gradients and surface normals, respectively, contribute to improvement of accuracy in an complementary fashion. Experimental results show that these two improvements enable to attain higher accuracy than the current state-of-the-arts, which is given by finer resolution reconstruction, for example, with small objects and object boundaries.


  • Open Access

    Deep Structured Energy-Based Image Inpainting

    by Fazil Altinel, Mete Ozay, Takayuki Okatani

    In this paper, we propose a structured image inpainting method employing an energy based model. In order to learn structural relationship between patterns observed in images and missing regions of the images, we employ an energy-based structured prediction method. The structural relationship is learned by minimizing an energy function which is defined by a simple convolutional neural network. The experimental results on various benchmark datasets show that our proposed method significantly outperforms the state-of-the-art methods which use Generative Adversarial Networks (GANs). We obtained 497.35 mean squared error (MSE) on the Olivetti face dataset compared to 833.0 MSE provided by the state-of-the-art method. Moreover, we obtained 28.4 dB peak signal to noise ratio (PSNR) on the SVHN dataset and 23.53 dB on the CelebA dataset, compared to 22.3 dB and 21.3 dB, provided by the state-of-the-art methods, respectively. The code is publicly available.

  • Encoding the local connectivity patterns of fMRI for cognitive task and state classification

    by Itir Onal Ertugrul, Mete Ozay, Fatos T Yarman Vural

    In this work, we propose a novel framework to encode the local connectivity patterns of brain, using Fisher vectors (FV), vector of locally aggregated descriptors (VLAD) and bag-of-words (BoW) methods. We first obtain local descriptors, called mesh arc descriptors (MADs) from fMRI data, by forming local meshes around anatomical regions, and estimating their relationship within a neighborhood. Then, we extract a dictionary of relationships, called brain connectivity dictionary by fitting a generative Gaussian mixture model (GMM) to a set of MADs, and selecting codewords at the mean of each component of the mixture. Codewords represent connectivity patterns among anatomical regions. We also encode MADs by VLAD and BoW methods using k-Means clustering. We classify cognitive tasks using the Human Connectome Project (HCP) task fMRI dataset and cognitive states using the Emotional Memory Retrieval (EMR). We train support vector machines (SVMs) using the encoded MADs. Results demonstrate that, FV encoding of MADs can be successfully employed for classification of cognitive tasks, and outperform VLAD and BoW representations. Moreover, we identify the significant Gaussians in mixture models by computing energy of their corresponding FV parts, and analyze their effect on classification accuracy. Finally, we suggest a new method to visualize the codewords of the learned brain connectivity dictionary.


  • Open Access

    Exploiting the Potential of Standard Convolutional Autoencoders for Image Restoration by Evolutionary Search

    by Masanori Suganuma, Mete Ozay, Takayuki Okatani

    Researchers have applied deep neural networks to image restoration tasks, in which they proposed various network architectures, loss functions, and training methods. In particular, adversarial training, which is employed in recent studies, seems to be a key ingredient to success. In this paper, we show that simple convolutional autoencoders (CAEs) built upon only standard network components, i.e., convolutional layers and skip connections, can outperform the state-of-the-art methods which employ adversarial training and sophisticated loss functions. The secret is to employ an evolutionary algorithm to automatically search for good architectures. Training optimized CAEs by minimizing the $\ell_2$ loss between reconstructed images and their ground truths using the ADAM optimizer is all we need. Our experimental results show that this approach achieves 27.8 dB peak signal to noise ratio (PSNR) on the CelebA dataset and 40.4 dB on the SVHN dataset, compared to 22.8 dB and 33.0 dB provided by the former state-of-the-art methods, respectively.

  • Open Access

    Optimization on Product Submanifolds of Convolution Kernels

    by Mete Ozay, Takayuki Okatani

    Recent advances in optimization methods used for training convolutional neural networks (CNNs) with kernels, which are normalized according to particular constraints, have shown remarkable success. This work introduces an approach for training CNNs using ensembles of joint spaces of kernels constructed using different constraints. For this purpose, we address a problem of optimization on ensembles of products of submanifolds (PEMs) of convolution kernels. To this end, we first propose three strategies to construct ensembles of PEMs in CNNs. Next, we expound their geometric properties (metric and curvature properties) in CNNs. We make use of our theoretical results by developing a geometry-aware SGD algorithm (G-SGD) for optimization on ensembles of PEMs to train CNNs. Moreover, we analyze convergence properties of G-SGD considering geometric properties of PEMs. In the experimental analyses, we employ G-SGD to train CNNs on Cifar-10, Cifar-100 and Imagenet datasets. The results show that geometric adaptive step size computation methods of G-SGD can improve training loss and convergence properties of CNNs. Moreover, we observe that classification performance of baseline CNNs can be boosted using G-SGD on ensembles of PEMs identified by multiple constraints.

  • Hierarchical multi-resolution mesh networks for brain decoding

    by Itir Onal Ertugrul, Mete Ozay, Fatos T. Yarman Vural

    Human brain is supposed to process information in multiple frequency bands. Therefore, we can extract diverse information from functional Magnetic Resonance Imaging (fMRI) data by processing it at multiple resolutions. We propose a framework, called Hierarchical Multi-resolution Mesh Networks (HMMNs), which establishes a set of brain networks at multiple resolutions of fMRI signal to represent the underlying cognitive process. Our framework, first, decomposes the fMRI signal into various frequency subbands using wavelet transform. Then, a brain network is formed at each subband by ensembling a set of local meshes. Arc weights of each local mesh are estimated by ridge regression. Finally, adjacency matrices of mesh networks obtained at different subbands are used to train classifiers in an ensemble learning architecture, called fuzzy stacked generalization (FSG). Our decoding performances on Human Connectome Project task-fMRI dataset reflect that HMMNs can successfully discriminate tasks with 99% accuracy, across 808 subjects. Diversity of information embedded in mesh networks of multiple subbands enables the ensemble of classifiers to collaborate with each other for brain decoding. The suggested HMMNs decode the cognitive tasks better than a single classifier applied to any subband. Also mesh networks have a better representation power compared to pairwise correlations or average voxel time series. Moreover, fusion of diverse information using FSG outperforms fusion with majority voting. We conclude that, fMRI data, recorded during a cognitive task, provide diverse information in multi-resolution mesh networks. Our framework fuses this complementary information and boosts the brain decoding performances obtained at individual subbands.


  • Open Access

    Truncating Wide Networks using Binary Tree Architectures

    by Yan Zhang, Mete Ozay, Shuohao Li, Takayuki Okatani

    Recent study shows that a wide deep network can obtain accuracy comparable to a deeper but narrower network. Compared to narrower and deeper networks, wide networks employ relatively less number of layers and have various important benefits, such that they have less running time on parallel computing devices, and they are less affected by gradient vanishing problems. However, the parameter size of a wide network can be very large due to use of large width of each layer in the network. In order to keep the benefits of wide networks meanwhile improve the parameter size and accuracy trade-off of wide networks, we propose a binary tree architecture to truncate architecture of wide networks by reducing the width of the networks. More precisely, in the proposed architecture, the width is continuously reduced from lower layers to higher layers in order to increase the expressive capacity of network with a less increase on parameter size. Also, to ease the gradient vanishing problem, features obtained at different layers are concatenated to form the output of our architecture. By employing the proposed architecture on a baseline wide network, we can construct and train a new network with same depth but considerably less number of parameters. In our experimental analyses, we observe that the proposed architecture enables us to obtain better parameter size and accuracy trade-off compared to baseline networks using various benchmark image classification datasets. The results show that our model can decrease the classification error of baseline from 20.43% to 19.22% on Cifar-100 using only 28% of parameters that baseline has. Code is available at https://github.com/ZhangVision/bitnet.

  • Open Access

    Design of Kernels in Convolutional Neural Networks for Image Classification

    by Zhun Sun, Mete Ozay, Takayuki Okatani

    Despite the effectiveness of Convolutional Neural Networks (CNNs) for image classification, our understanding of the relationship between shape of convolution kernels and learned representations is limited. In this work, we explore and employ the relationship between shape of kernels which define Receptive Fields (RFs) in CNNs for learning of feature representations and image classification. For this purpose, we first propose a feature visualization method for visualization of pixel-wise classification score maps of learned features. Motivated by our experimental results, and observations reported in the literature for modeling of visual systems, we propose a novel design of shape of kernels for learning of representations in CNNs. In the experimental results, we achieved a state-of-the-art classification performance compared to a base CNN model [28] by reducing the number of parameters and computational time of the model using the ILSVRC-2012 dataset [24]. The proposed models also outperform the state-of-the-art models employed on the CIFAR-10/100 datasets [12] for image classification. Additionally, we analyzed the robustness of the proposed method to occlusion for classification of partially occluded images compared with the state-of-the-art methods. Our results indicate the effectiveness of the proposed approach. The code is available in github.com/minogame/caffe-qhconv.

  • Open Access

    Optimization on Submanifolds of Convolution Kernels in CNNs

    by Mete Ozay, Takayuki Okatani

    Kernel normalization methods have been employed to improve robustness of optimization methods to reparametrization of convolution kernels, covariate shift, and to accelerate training of Convolutional Neural Networks (CNNs). However, our understanding of theoretical properties of these methods has lagged behind their success in applications. We develop a geometric framework to elucidate underlying mechanisms of a diverse range of kernel normalization methods. Our framework enables us to expound and identify geometry of space of normalized kernels. We analyze and delineate how state-of-the-art kernel normalization methods affect the geometry of search spaces of the stochastic gradient descent (SGD) algorithms in CNNs. Following our theoretical results, we propose a SGD algorithm with assurance of almost sure convergence of the methods to a solution at single minimum of classification loss of CNNs. Experimental results show that the proposed method achieves state-of-the-art performance for major image classification benchmarks with CNNs.