https://scholarworks.unist.ac.kr/handle/201301/30 Wed, 08 Apr 2026 00:26:44 GMT 2026-04-08T00:26:44Z https://scholarworks.unist.ac.kr/handle/201301/88288 Title: Reliable Input Attribution through Path Modeling in Model Structure and Input Space Author(s): Lim, Seongwoo Major: Department of Computer Science and Engineering Thu, 31 Jul 2025 15:00:00 GMT https://scholarworks.unist.ac.kr/handle/201301/88288 2025-07-31T15:00:00Z https://scholarworks.unist.ac.kr/handle/201301/88289 Title: Resource-Efficient and Convergence-Preserving DNN Training Parallelism Techniques Author(s): Yun, Gyeongchan Major: Department of Computer Science and Engineering Thu, 31 Jul 2025 15:00:00 GMT https://scholarworks.unist.ac.kr/handle/201301/88289 2025-07-31T15:00:00Z https://scholarworks.unist.ac.kr/handle/201301/86539 Title: Approximation Algorithms in Hyperbolic Spaces with Constant Additive Error Author(s): Park, Eunku Abstract: Previous work in computational geometry has focused on Euclidean spaces, where the distance between two points p and q is given by the L2 norm of the vector pq. In this thesis, we consider algorithmic problems in hyperbolic spaces, using different metrics. Our motivation is that, even though Euclidean spaces are suitable for many applications, there has been recently some interest in hyperbolic spaces for applications in computer networks, artificial neural networks, and computer vision. In the main part of this thesis, we present an embedding of the D-dimensional Poincaré half-space into a discrete hyperbolic space that is based on a binary tiling of the upper half-space. Based on this embedding, we obtain an embedding of any finite subset of the Poincaré half-space into a graph metric with a linear number of edges, and a constant additive error. We extend this result to obtain a spanner with a linear number of Steiner points, and a linear number of edges, still with a constant additive error. Both of these constructions can be made in near-linear time. Finally, we show how to construct an approximate Voronoi diagram in an hyperbolic space, with constant additive error. It yields a data structure for answering approximate near-neighbor searching queries in logarithmic time, with constant additive error. The approximate Voronoi diagram and the associated data structure can be computed in near-linear time. Major: Department of Computer Science and Engineering Fri, 31 Jan 2025 15:00:00 GMT https://scholarworks.unist.ac.kr/handle/201301/86539 2025-01-31T15:00:00Z https://scholarworks.unist.ac.kr/handle/201301/84231 Title: Learning to Achieve Goals via Curriculum and Hierarchical Reinforcement Learning Author(s): Lee, Kyowoon Abstract: This thesis explores methods for learning to achieve goals through curriculum and hierarchical reinforcement learning (RL). As intelligent agents increasingly engage with complex, real-world environments, their ability to autonomously learn and adapt skills without extensive human intervention becomes crucial. Traditional RL approaches, however, often depend on hand- engineered curricula and reward functions, limiting their efficiency and adaptability. This thesis addresses these limitations by introducing novel approaches that leverage curriculum learning and hierarchical structures to improve skill acquisition and deployment in goal-conditioned RL. The aforementioned problems are both important and challenging. As agents increasingly engage in complex and dynamic environments, the need for autonomous skill acquisition and adaptive deployment becomes crucial. Without the ability to learn and adapt without human intervention, the scalability and real-world applicability of RL agents are severely restricted. Efficient goal achievement is further hindered by the complexity of environmental interactions and the diverse requirements of different tasks. Therefore, it is essential to develop methods allowing agents to autonomously acquire a diverse set of skills and adapt them dynamically to various situations. This thesis aims to enhance goal-conditioned reinforcement learning by integrating curricu- lum and hierarchical learning methods, enabling agents to autonomously and efficiently achieve goals in complex, real-world environments. The contributions include the development of Varia- tional Curriculum Reinforcement Learning (VCRL) and Value Uncertainty Variational Curricu- lum (VUVC) for effective unsupervised goal achievement and curriculum learning, as well as the creation of a hierarchical learning framework for adaptive and explainable skill deployment. Major: Department of Computer Science and Engineering Wed, 31 Jul 2024 15:00:00 GMT https://scholarworks.unist.ac.kr/handle/201301/84231 2024-07-31T15:00:00Z