Kamer Kaya

Deniz Elbek, Kamer Kaya

Modern GPUs have Tensor Cores (TCs) capable of extremely high-throughput matrix operations, yet graph algorithms remain difficult to accelerate because of their irregular and data-dependent execution patterns. This work presents BLEST, a TC-accelerated framework that reformulates Breadth-First Search (BFS) as a bit-level sparse matrix-vector computation while addressing the load imbalance, memory inefficiency, and synchronization overheads that limit prior approaches. BLEST introduces Binarized Virtual Slice Sets (BVSS), a graph representation that partitions work into balanced warp-level units and schedules only frontier-relevant regions of the graph. It further employs an optimized TC layout that maps neighbour checks onto binary MMA instructions without wasted outputs, reducing the number of required MMA calls by 8$\times$ compared with prior layouts. To mitigate atomic and cache bottlenecks, BLEST incorporates a lazy vertex-update scheme. We revisit the switching terminology for BFS and propose a mechanism that dynamically transitions from TCs to CUDA cores when it becomes more efficient. We also extend BLEST to multi-source BFS and closeness centrality workloads. Finally, we introduce a scalable graph reordering method that improves compression for scale-free-like graphs, while using RCM to improve locality for others. Across a broad set of real-world graphs, BLEST achieves average speedups of 22.0$\times$, 7.7$\times$, 8.1$\times$, and 5.9$\times$ over GAP, Gunrock, GSWITCH, and BerryBees, respectively, establishing a new BFS baseline on GPUs. Thanks to its high performance, BLEST can compute the exact closeness centralities of 65.6M vertices in a social network with 3.6B edges in an hour using 100 H100 GPUs.

Mehmet Can Şakiroğlu, H. Altay Güvenir, Kamer Kaya

Generating multiple-choice questions (MCQs) with difficulty estimation remains challenging in automated MCQ-generation systems used in adaptive, AI-assisted education. This study proposes a novel methodology for generating MCQs with difficulty estimation from the input documents by utilizing knowledge graphs (KGs) and large language models (LLMs). Our approach uses an LLM to construct a KG from input documents, from which MCQs are then systematically generated. Each MCQ is generated by selecting a node from the KG as the key, sampling a related triple or quintuple -- optionally augmented with an extra triple -- and prompting an LLM to generate a corresponding stem from these graph components. Distractors are then selected from the KG. For each MCQ, nine difficulty signals are computed and combined into a unified difficulty score using a data-driven approach. Experimental results demonstrate that our method generates high-quality MCQs whose difficulty estimation is interpretable and aligns with human perceptions. Our approach improves automated MCQ generation by integrating structured knowledge representations with LLMs and a data-driven difficulty estimation model.

Ahmet Yasin Aytar, Kemal Kilic, Kamer Kaya

In the rapidly evolving field of data science, efficiently navigating the expansive body of academic literature is crucial for informed decision-making and innovation. This paper presents an enhanced Retrieval-Augmented Generation (RAG) application, an artificial intelligence (AI)-based system designed to assist data scientists in accessing precise and contextually relevant academic resources. The AI-powered application integrates advanced techniques, including the GeneRation Of BIbliographic Data (GROBID) technique for extracting bibliographic information, fine-tuned embedding models, semantic chunking, and an abstract-first retrieval method, to significantly improve the relevance and accuracy of the retrieved information. This implementation of AI specifically addresses the challenge of academic literature navigation. A comprehensive evaluation using the Retrieval-Augmented Generation Assessment System (RAGAS) framework demonstrates substantial improvements in key metrics, particularly Context Relevance, underscoring the system's effectiveness in reducing information overload and enhancing decision-making processes. Our findings highlight the potential of this enhanced Retrieval-Augmented Generation system to transform academic exploration within data science, ultimately advancing the workflow of research and innovation in the field.

Kemal Mutluergil, Deniz Elbek, Kamer Kaya et al.

Homomorphic encryption (HE) enables computation over encrypted data but incurs a substantial overhead. For sparse-matrix vector multiplication, the widely used Halevi and Shoup (2014) scheme has a cost linear in the number of occupied cyclic diagonals, which may be many due to the irregular nonzero pattern of the matrix. In this work, we study how to permute the rows and columns of a sparse matrix so that its nonzeros are packed into as few cyclic diagonals as possible. We formalise this as the two-dimensional diagonal packing problem (2DPP), introduce the two-dimensional circular bandsize metric, and give an integer programming formulation that yields optimal solutions for small instances. For large matrices, we propose practical ordering heuristics that combine graph-based initial orderings - based on bandwidth reduction, anti-bandwidth maximisation, and spectral analysis - and an iterative-improvement-based optimization phase employing 2OPT and 3OPT swaps. We also introduce a dense row/column elimination strategy and an HE-aware cost model that quantifies the benefits of isolating dense structures. Experiments on 175 sparse matrices from the SuiteSparse collection show that our ordering-optimisation variants can reduce the diagonal count by $5.5\times$ on average ($45.6\times$ for one instance). In addition, the dense row/column elimination approach can be useful for cases where the proposed permutation techniques are not sufficient; for instance, in one case, the additional elimination helped to reduce the encrypted multiplication cost by $23.7\times$ whereas without elimination, the improvement was only $1.9\times$.

Ceren Yildirim, Kamer Kaya, Sinan Yildirim et al.

We propose a new framework for Bayesian estimation of differential privacy, incorporating evidence from multiple membership inference attacks (MIA). Bayesian estimation is carried out via a Markov chain Monte Carlo (MCMC) algorithm, named MCMC-DP-Est, which provides an estimate of the full posterior distribution of the privacy parameter (e.g., instead of just credible intervals). Critically, the proposed method does not assume that privacy auditing is performed with the most powerful attack on the worst-case (dataset, challenge point) pair, which is typically unrealistic. Instead, MCMC-DP-Est jointly estimates the strengths of MIAs used and the privacy of the training algorithm, yielding a more cautious privacy analysis. We also present an economical way to generate measurements for the performance of an MIA that is to be used by the MCMC method to estimate privacy. We present the use of the methods with numerical examples with both artificial and real data.

Amro Alabsi Aljundi, Taha Atahan Akyıldız, Kamer Kaya

Graphs are ubiquitous, and they can model unique characteristics and complex relations of real-life systems. Although using machine learning (ML) on graphs is promising, their raw representation is not suitable for ML algorithms. Graph embedding represents each node of a graph as a d-dimensional vector which is more suitable for ML tasks. However, the embedding process is expensive, and CPU-based tools do not scale to real-world graphs. In this work, we present GOSH, a GPU-based tool for embedding large-scale graphs with minimum hardware constraints. GOSH employs a novel graph coarsening algorithm to enhance the impact of updates and minimize the work for embedding. It also incorporates a decomposition schema that enables any arbitrarily large graph to be embedded with a single GPU. As a result, GOSH sets a new state-of-the-art in link prediction both in accuracy and speed, and delivers high-quality embeddings for node classification at a fraction of the time compared to the state-of-the-art. For instance, it can embed a graph with over 65 million vertices and 1.8 billion edges in less than 30 minutes on a single GPU.

Taha Atahan Akyildiz, Amro Alabsi Aljundi, Kamer Kaya

A significant portion of the data today, e.g, social networks, web connections, etc., can be modeled by graphs. A proper analysis of graphs with Machine Learning (ML) algorithms has the potential to yield far-reaching insights into many areas of research and industry. However, the irregular structure of graph data constitutes an obstacle for running ML tasks on graphs such as link prediction, node classification, and anomaly detection. Graph embedding is a compute-intensive process of representing graphs as a set of vectors in a d-dimensional space, which in turn makes it amenable to ML tasks. Many approaches have been proposed in the literature to improve the performance of graph embedding, e.g., using distributed algorithms, accelerators, and pre-processing techniques. Graph coarsening, which can be considered a pre-processing step, is a structural approximation of a given, large graph with a smaller one. As the literature suggests, the cost of embedding significantly decreases when coarsening is employed. In this work, we thoroughly analyze the impact of the coarsening quality on the embedding performance both in terms of speed and accuracy. Our experiments with a state-of-the-art, fast graph embedding tool show that there is an interplay between the coarsening decisions taken and the embedding quality.

Stefan Wagner, Asim Abdulkhaleq, Kamer Kaya et al.

Background: Code cloning - copying and reusing pieces of source code - is a common phenomenon in software development in practice. There have been several empirical studies on the effects of cloning, but there are contradictory results regarding the connection of cloning and faults. Objective: Our aim is to clarify the relationship between code clones and faults. In particular, we focus on inconsistent (or type-3) clones in this work. Method: We conducted a case study with TWT GmbH where we detected the code clones in three Java systems, set them into relation to information from issue tracking and version control and interviewed three key developers. Results: Of the type-3 clones, 17 % contain faults. Developers modified most of the type-3 clones simultaneously and thereby fixed half of the faults in type-3 clones consistently. Type-2 clones with faults all evolved to fixed type-3 clones. Clone length is only weakly correlated with faultiness. Conclusion: There are indications that the developers in two cases have been aware of clones. It might be a reason for the weak relationship between type-3 clones and faults. Hence, it seems important to keep developers aware of clones, potentially with new tool support. Future studies need to investigate if the rate of faults in type-3 clones justifies using them as cues in defect detection.

Oguzhan Ersoy, Kamer Kaya, Kerem Kaskaloglu

A recent work of Harn and Fuyou presents the first multilevel (disjunctive) threshold secret sharing scheme based on the Chinese Remainder Theorem. In this work, we first show that the proposed method is not secure and also fails to work with a certain natural setting of the threshold values on compartments. We then propose a secure scheme that works for all threshold settings. In this scheme, we employ a refined version of Asmuth-Bloom secret sharing with a special and generic Asmuth-Bloom sequence called the {\it anchor sequence}. Based on this idea, we also propose the first multilevel conjunctive threshold secret sharing scheme based on the Chinese Remainder Theorem. Lastly, we discuss how the proposed schemes can be used for multilevel threshold function sharing by employing it in a threshold RSA cryptosystem as an example.

Kamer Kaya, Figen Öztoprak, Ş. İlker Birbil et al.

We propose HAMSI (Hessian Approximated Multiple Subsets Iteration), which is a provably convergent, second order incremental algorithm for solving large-scale partially separable optimization problems. The algorithm is based on a local quadratic approximation, and hence, allows incorporating curvature information to speed-up the convergence. HAMSI is inherently parallel and it scales nicely with the number of processors. Combined with techniques for effectively utilizing modern parallel computer architectures, we illustrate that the proposed method converges more rapidly than a parallel stochastic gradient descent when both methods are used to solve large-scale matrix factorization problems. This performance gain comes only at the expense of using memory that scales linearly with the total size of the optimization variables. We conclude that HAMSI may be considered as a viable alternative in many large scale problems, where first order methods based on variants of stochastic gradient descent are applicable.

Onur Küçüktunç, Erik Saule, Kamer Kaya et al.

Literature search is arguably one of the most important phases of the academic and non-academic research. The increase in the number of published papers each year makes manual search inefficient and furthermore insufficient. Hence, automatized methods such as search engines have been of interest in the last thirty years. Unfortunately, these traditional engines use keyword-based approaches to solve the search problem, but these approaches are prone to ambiguity and synonymy. On the other hand, bibliographic search techniques based only on the citation information are not prone to these problems since they do not consider textual similarity. For many particular research areas and topics, the amount of knowledge to humankind is immense, and obtaining the desired information is as hard as looking for a needle in a haystack. Furthermore, sometimes, what we are looking for is a set of documents where each one is different than the others, but at the same time, as a whole we want them to cover all the important parts of the literature relevant to our search. This paper targets the problem of result diversification in citation-based bibliographic search. It surveys a set of techniques which aim to find a set of papers with satisfactory quality and diversity. We enhance these algorithms with a direction-awareness functionality to allow the users to reach either old, well-cited, well-known research papers or recent, less-known ones. We also propose a set of novel techniques for a better diversification of the results. All the techniques considered are compared by performing a rigorous experimentation. The results show that some of the proposed techniques are very successful in practice while performing a search in a bibliographic database.

Onur Küçüktunç, Erik Saule, Kamer Kaya et al.

The literature search has always been an important part of an academic research. It greatly helps to improve the quality of the research process and output, and increase the efficiency of the researchers in terms of their novel contribution to science. As the number of published papers increases every year, a manual search becomes more exhaustive even with the help of today's search engines since they are not specialized for this task. In academics, two relevant papers do not always have to share keywords, cite one another, or even be in the same field. Although a well-known paper is usually an easy pray in such a hunt, relevant papers using a different terminology, especially recent ones, are not obvious to the eye. In this work, we propose paper recommendation algorithms by using the citation information among papers. The proposed algorithms are direction aware in the sense that they can be tuned to find either recent or traditional papers. The algorithms require a set of papers as input and recommend a set of related ones. If the user wants to give negative or positive feedback on the suggested paper set, the recommendation is refined. The search process can be easily guided in that sense by relevance feedback. We show that this slight guidance helps the user to reach a desired paper in a more efficient way. We adapt our models and algorithms also for the venue and reviewer recommendation tasks. Accuracy of the models and algorithms is thoroughly evaluated by comparison with multiple baselines and algorithms from the literature in terms of several objectives specific to citation, venue, and reviewer recommendation tasks. All of these algorithms are implemented within a publicly available web-service framework (http://theadvisor.osu.edu/) which currently uses the data from DBLP and CiteSeer to construct the proposed citation graph.