News – Page 4

August 5, 2025Dr. Algın and his collaborators’ MRI robotic brain intervention system project has received R&D support from the Turkish Institutes of Health (TÜSEB) and the German R&D companies MARVIS and NORAS.

Recent research by Prof. Dr. Oktay Algın and his collaborators has led to the development of MRI-compatible brain biopsy and ablation techniques. These innovations have received support from the Turkish Institutes of…

May 6, 2025Visibility Graphs in Natural Language Processing

Happy to share our latest paper titled "Emotion Classification with Visibility Graphs" published in IEEE Signal Processing Letters! Our novel approach introduces a visibility graph-based sequential modeling to encapsulate global relationships within…

Happy to share our latest paper titled “Emotion Classification with Visibility Graphs” published in IEEE Signal Processing Letters! Our novel approach introduces a visibility graph-based sequential modeling to encapsulate global relationships within texts beyond the reach of fixed context windows of transformers. Being a modular graph-based extension to any transformer-based model, our proposed approach enhances the transformer-based models as a plug-in model, enabling them to outperform the state-of-the-art methods on the common largescale benchmarks: SemEval2018 and GoEmotions. Check out the paper for more details! Paper: https://ieeexplore.ieee.org/document/10985783 Code: https://github.com/koc-lab/EmoVis Abstract: Transformers have gained prominence in natural language processing due to their representational capabilities and performances. Transformers process natural language as a sequence on finite context windows; however, global relationships among words beyond these windows cannot be completely modeled via sequence processing only. Graph neural network (GNN) based models have been proposed to alleviate this problem, as they provide geometric extensions to neural networks, enabling models to learn associations within a text. However, regular graph-based methods ignore the sequential nature of underlying texts. In this paper, we propose EmoVis, the first generic graph-based neural network that utilizes visibility graphs, which converts classical time-series information to graph representations. We cast the problem as an emotion classification task, enabling the proposed model to learn associations between the labels and words in a sentence. Moreover, EmoVis can be used as a highly modular graph-based extension to any transformer-based model, significantly improving their performance and learning capabilities in various languages. We experimentally show that EmoVis enables transformer-based models to outperform the state-of-the-art baselines across three diverse datasets in different languages in the SemEval2018 competition datasets and the GoEmotions dataset.

February 24, 2025Joint Time-Vertex Fractional Fourier Transform

Excited to share our latest research published in Signal Processing, introducing the Joint Time-Vertex Fractional Fourier Transform (JFRT), a novel framework that extends traditional joint time-vertex analysis into the fractional domain. By…

Excited to share our latest research published in Signal Processing, introducing the Joint Time-Vertex Fractional Fourier Transform (JFRT), a novel framework that extends traditional joint time-vertex analysis into the fractional domain. By integrating fractional orders in both time and graph domains, JFRT not only generalizes existing Fourier-based methods but also delivers enhanced performance in tasks like denoising, filtering, and clustering. Key Highlights: ● A unified approach for fractional analysis of joint time-vertex signals. ● Theoretical guarantees including index additivity, reversibility, and reduction to identity. ● A Tikhonov regularization-based denoising method that robustly handles noisy data. ● Comprehensive numerical experiments demonstrating improved results over state-of-the-art techniques. In summary, our new JFRT method sets a new benchmark for joint time-vertex signal analysis by combining enhanced flexibility, efficiency, and robustness in both time and graph domains. We’re excited about the potential applications of JFRT in advanced graph signal processing and beyond! #SignalProcessing #GraphSignalProcessing #FractionalFourierTransform Paper: https://doi.org/10.1016/j.sigpro.2025.109944 Abstract: Graph signal processing (GSP) facilitates the analysis of high-dimensional data on non-Euclidean domains by utilizing graph signals defined on graph vertices. In addition to static data, each vertex can provide continuous time-series signals, transforming graph signals into time-series signals on each vertex. The joint time-vertex Fourier transform (JFT) framework offers spectral analysis capabilities to analyze these joint time-vertex signals. Analogous to the fractional Fourier transform (FRT) extending the ordinary Fourier transform (FT), we introduce the joint time-vertex fractional Fourier transform (JFRT) as a generalization of JFT. The JFRT enables fractional analysis for joint time-vertex processing by extending Fourier analysis to fractional orders in both temporal and vertex domains. We theoretically demonstrate that JFRT generalizes JFT and maintains properties such as index additivity, reversibility, reduction to identity, and unitarity for specific graph topologies. Additionally, we derive Tikhonov regularization-based denoising in the JFRT domain, ensuring robust and well-behaved solutions. Comprehensive numerical experiments on synthetic and real-world datasets highlight the effectiveness of JFRT in denoising and clustering tasks that outperform state-of-the-art approaches.

February 21, 2025New studies by Dr. Algin and his team on Diffusion Tensor Imaging and MR Cisternography have been published in SCI journals.

Oktay Algin and his colleagues conducted two studies published in two international scientific journals. The names of these studies are ‘Evaluation of the Glymphatic System in Rabbits Using Gadobutrol-Enhanced MR Cisternography With T1 and…

January 29, 2025Semantic Communication Over Channels with Insertions, Deletions, and Substitutions

Happy to share the first semantic communication framework on insertion-deletion-substitution (IDS) channels as published in our latest paper titled "Semantic Communication over Channels with Insertions, Deletions, and Substitutions" published in IEEE Communications…

Happy to share the first semantic communication framework on insertion-deletion-substitution (IDS) channels as published in our latest paper titled “Semantic Communication over Channels with Insertions, Deletions, and Substitutions” published in IEEE Communications Letters! A significant class of binary input channels are prone to synchronization errors, which are modeled as IDS channels. IDS channels occur in many scenarios like wireless communications, file synchronization, bit-patterned media, racetrack memories, and DNA storage. In our paper, we propose the first joint semantic transmission scheme for binary channels with synchronization errors, a novel training strategy that utilizes a pre-trained error correction network in the joint-source channel coding schemes with a transfer learning approach. Our proposed architecture consists of a three-stage training algorithm that combines, for the first time, gated recurrent unit (GRU) networks for marker detection, transformer-based semantic communication for continuous latent space, and lookup-free quantization for binarized latent space optimization, specifically tailored to IDS channels. Check out the paper for more details! Paper: https://ieeexplore.ieee.org/abstract/document/10852305 Code: https://github.com/koc-lab/deepjsoc Abstract: We present deep joint source-outer channel coding (DeepJSOC), an end-to-end deep learning-based semantic communication architecture designed for channels with insertions, deletions, and substitutions (IDS). We propose a three-stage training algorithm that combines, for the first time, gated recurrent unit (GRU) networks for marker detection, transformer-based semantic communication for continuous latent space, and lookup-free quantization for binarized latent space optimization, specifically tailored to IDS channels. The proposed DeepJSOC is the first to integrate deep learning-based error correction networks into joint-source channel coding schemes for binary channels with synchronization errors. We demonstrate the effectiveness of DeepJSOC by experiments, achieving significant improvements over existing methods in text transmission.

August 29, 2024Sememe Based Semantic Communications

Happy to share our latest paper titled "Sememe Based Semantic Communications" published in IEEE Communications Letters! We introduced a concept in linguistics called sememes to the domain of semantic communications. Sememes are…

Happy to share our latest paper titled “Sememe Based Semantic Communications” published in IEEE Communications Letters! We introduced a concept in linguistics called sememes to the domain of semantic communications. Sememes are the smallest and indivisible semantic units of word meaning. A predefined set of sememes is theoretically considered “the periodic table” of meaning in a natural language. Expressing words in terms of combinations of a limited number of sememes is essential for interpreting the intrinsic semantics of a word, which is an essential step towards “defining the meaning behind bits”, which is central importance in semantic communications. We proposed SememeSC, which adopts a novel semantic encoding method based on sememes and a semantic decoding method based on deep learning (DL) for text transmission. Our extensive experimentation shows that SememeSC outperforms baselines in additive white Gaussian noise (AWGN) and Rayleigh fading channels. Check out the paper for more details! Paper: https://ieeexplore.ieee.org/document/10648841 Code: https://github.com/koc-lab/sememesc Abstract: Semantic communication, a paradigm concentrating on correctly transmitting underlying semantic information instead of bit sequences, has proved effective. Deep learning (DL) enabled methods are mainly used with basic natural language processing (NLP) techniques for the semantic communication of texts. However, most of the previous work approaches the problem by treating text as a generic continuum of sequences of textual information without leveraging underlying intrinsic and advanced linguistic properties of natural languages. A prime example of such linguistic features is the sememes, the smallest and indivisible semantic units of textual information (analogous to the Periodic Table of Matter). This paper proposes Sememe-based Semantic Communications (SememeSC), a semantic communication paradigm that utilizes sememe knowledge in natural languages. We provided experimental results verifying that the proposed SememeSC performs superior to baselines in additive white Gaussian noise (AWGN) and Rayleigh fading channels.

August 14, 2024VISPool: Enhancing Transformer Encoders with Vector Visibility Graph Neural Networks

We are happy to announce that we demonstrated our latest work titled “VISPool: Enhancing Transformer Encoders with Vector Visibility Graph Neural Networks” at ACL 2024, Bangkok/Thailand! In our #ACL2024 paper, we…

We are happy to announce that we demonstrated our latest work titled “VISPool: Enhancing Transformer Encoders with Vector Visibility Graph Neural Networks” at ACL 2024, Bangkok/Thailand! In our #ACL2024 paper, we explore the powerful combination of transformers and graph neural networks (GNNs) to push the boundaries of natural language processing (NLP). While transformers have revolutionized NLP, integrating GNNs presents a promising avenue, though existing methods often struggle with limitations of transductive approach, scalability and complex text processing. To address these challenges, we introduce a novel dynamic graph construction method for text documents, leveraging vector visibility graphs (VVGs) generated from transformer outputs. Our proposed model, the visibility pooler (VISPool), seamlessly integrates VVG convolutional networks into transformer pipelines. Notably, VISPool outperforms baselines on the GLUE benchmark with fewer trainable parameters, highlighting the effectiveness of our approach in enhancing transformers with GNNs. For more details, check out the full paper and source code! #ACL2024 #NLP #Transformers #GNNs #AI #MachineLearning #ACL #AI Paper: https://aclanthology.org/2024.findings-acl.149 Code: https://github.com/koc-lab/vispool Abstract: The emergence of transformers has revolutionized natural language processing (NLP), as evidenced in various NLP tasks. While graph neural networks (GNNs) show recent promise in NLP, they are not standalone replacements for transformers. Rather, recent research explores combining transformers and GNNs. Existing GNN-based approaches rely on static graph construction methods requiring excessive text processing, and most of them are not scalable with the increasing document and word counts. We address these limitations by proposing a novel dynamic graph construction method for text documents based on vector visibility graphs (VVGs) generated from transformer output. Then, we introduce visibility pooler (VISPool), a scalable model architecture that seamlessly integrates VVG convolutional networks into transformer pipelines. We evaluate the proposed model on the General Language Understanding Evaluation (GLUE) benchmark datasets. VISPool outperforms the baselines with less trainable parameters, demonstrating the viability of the visibility-based graph construction method for enhancing transformers with GNNs.

August 9, 2024Graph Fractional Fourier Transform: A Unified Theory

We are pleased to announce that our paper on generalizing the fractional Fourier transform to the graph domain titled "Graph Fractional Fourier Transform: A Unified Theory" published in IEEE Transactions on Signal…

We are pleased to announce that our paper on generalizing the fractional Fourier transform to the graph domain titled “Graph Fractional Fourier Transform: A Unified Theory” published in IEEE Transactions on Signal Processing! Highlights of our contributions include:

A rigorous extension of the fractional power-based definition of GFRFT to support any graph structure and transform order.
Introduction of a novel hyper-differential operator-based GFRFT definition, ensuring consistency with classical FRFT theory and enabling faster computations for large graphs.
A procedure to optimally select the transform order via learning, making it possible to apply GFRFT effectively within neural networks and machine learning applications.
Comprehensive experimental validation demonstrating the equivalence of GFRFT definitions and the advantages over GFT and other GSP methods in tasks such as denoising, classification, and sampling.

We believe this unified theory for GFRFT will open up new research avenues and provide a powerful tool for analyzing graph-based data. Check out the paper for more details! #Research #SignalProcessing #MachineLearning #GraphSignalProcessing #FractionalFourierTransform #IEEE Paper: https://doi.org/10.1109/TSP.2024.3439211 Code: https://github.com/koc-lab/gfrft-unified Abstract: The fractional Fourier transform (FRFT) parametrically generalizes the Fourier transform (FT) by a transform order, representing signals in intermediate time-frequency domains. The FRFT has multiple but equivalent definitions, including the fractional power of FT, time-frequency plane rotation, hyper-differential operator, and many others, each offering benefits like derivational ease and computational efficiency. Concurrently, graph signal processing (GSP) extends traditional signal processing to data on irregular graph structures, enabling concepts like sampling, filtering, and Fourier transform for graph signals. The graph fractional Fourier transform (GFRFT) has been recently extended to the GSP domain. However, this extension only generalizes the fractional power definition of FRFT based on specific graph structures with limited transform order range. Ideally, the GFRFT extension should be consistent with as many alternative definitions as possible. This paper first provides a rigorous fractional power-based GFRFT definition that supports any graph structure and transform order. Then, we introduce the novel hyper-differential operator-based GFRFT definition, allowing faster forward and inverse transform matrix computations on large graphs. Through the proposed definition, we derive a novel approach to select the transform order by learning the optimal value from data. Furthermore, we provide treatments of the core GSP concepts, such as bandlimitedness, filters, and relations to the other transforms in the context of GFRFT. Finally, with comprehensive experiments, including denoising, classification, and sampling tasks, we demonstrate the equivalence of parallel definitions of GFRFT, learnability of the transform order, and the benefits of GFRFT over GFT and other GSP methods.

August 7, 2024Text-RGNNs: Relational Modeling for Heterogeneous Text Graphs

Happy to share our latest paper titled "Text-RGNNs: Relational Modeling for Heterogeneous Text Graphs" published in IEEE Signal Processing Letters! Building on the foundational Text-graph convolutional Network (TextGCN), which represents corpus with…

Happy to share our latest paper titled “Text-RGNNs: Relational Modeling for Heterogeneous Text Graphs” published in IEEE Signal Processing Letters! Building on the foundational Text-graph convolutional Network (TextGCN), which represents corpus with heterogeneous text graphs, we addressed a key limitation: GCNs are inherently designed to operate within homogeneous graphs, potentially limiting their performance. To overcome this, based on relational graph neural networks (RGNNs), we introduced Text-RGNNs, assigning dedicated weight matrices to each relation using heterogeneous GNNs for more nuanced and effective modeling. Our extensive experimentation shows that Text-RGNNs significantly outperform existing state-of-the-art models, especially with complete labeled nodes and minimal labeled training data, achieving up to 10.61% improvement in evaluation metrics! Check out the paper for more details! Paper: https://doi.org/10.1109/LSP.2024.3433568 Code: https://github.com/koc-lab/text-rgnn Abstract: Text-graph convolutional network (TextGCN) is the fundamental work representing corpus with heterogeneous text graphs. Its innovative application of GCNs for text classification has garnered widespread recognition. However, GCNs are inherently designed to operate within homogeneous graphs, potentially limiting their performance. To address this limitation, we present Text Relational Graph Neural Networks (Text-RGNNs), which offer a novel methodology by assigning dedicated weight matrices to each relation within the graph by using heterogeneous GNNs. This approach leverages RGNNs, enabling nuanced and compelling modeling of relationships inherent in the heterogeneous text graphs, ultimately resulting in performance enhancements. We present a theoretical framework for the relational modeling of GNNs for text classification within the context of document classification and demonstrate its effectiveness through extensive experimentation on benchmark datasets. Conducted experiments reveal that Text-RGNNs outperform the existing state-of-the-art in scenarios with complete labeled nodes and minimal labeled training data proportions by incorporating relational modeling into heterogeneous text graphs. Text-RGNNs outperform the second-best models by up to 10.61% for the corresponding evaluation metric.

May 15, 2024UMRAM presentations at ISMRM 2024

UMRAM had broad participation in the 2024 ISMRM & ISMRT Annual Meeting & Exhibition held 04-09 May 2024 in Singapore. Two educational lectures as well as five oral, one power…

May 15, 2024Prof. Tolga Çukur Named Fellow of ISMRM

Prof. Tolga Çukur of the Department of Electrical and Electronics Engineering and director of UMRAM has been named a fellow of International Society for Magnetic Resonance in Medicine (ISMRM). ISMRM is an…

May 6, 2024Wiener Filtering in Joint Time-Vertex Fractional Fourier Domains

We are excited to share our latest work on joint time-vertex signal processing published in IEEE Signal Processing Letters! In this paper, we explore the complexities of time-varying graph signals and how…

We are excited to share our latest work on joint time-vertex signal processing published in IEEE Signal Processing Letters! In this paper, we explore the complexities of time-varying graph signals and how they can be more efficiently processed using the joint time-vertex framework. Traditionally, separating signal from noise in these structures has posed significant challenges. However, our research introduces an innovative approach using the fractional Fourier domain, which has proven to achieve lower error rates compared to conventional methods. 🔍 Key Highlights:

We solve the optimal Wiener filtering problem within a new joint time-vertex fractional Fourier framework.
Our theoretical analysis and numerical experiments provide a novel perspective and demonstrate superior performance over existing filtering techniques.

The proposed method not only enhances our ability to handle time-varying graph signals but also opens up new possibilities for more sophisticated data analysis in network structures by using parametered transformations. #jointtimevertex #jtv #wiener #FourierTransform #FractionalFourierTransform #signalprocessing Paper: https://ieeexplore.ieee.org/document/10518099 Code: https://github.com/koc-lab/jfrt-optimal Abstract: Graph signal processing (GSP) uses network structures to analyze and manipulate interconnected signals. Graph signals can also be time-varying. The established joint time-vertex processing framework and corresponding joint time-vertex Fourier transform provide a basis to endeavor such time-varying graph signals. The optimal Wiener filtering problem has been deliberated within the joint time-vertex framework. However, the ordinary Fourier domain is only sometimes optimal for separating the signal and noise; one can achieve lower error rates in a fractional Fourier domain. In this paper, we solve the optimal Wiener filtering problem in the joint time-vertex fractional Fourier domains. We provide a theoretical analysis and numerical experiments with comprehensive comparisons to existing filtering approaches for time-varying graph signals to demonstrate the advantages of our approach.