EE7001/EE8001: CS Seminar, 114-2, Spring 2026

Course Staffs

• Instructor: Shih-Han Hung
• TA: George Cheng

Schedule and Grading

• Location: MD-231 (明達231)
• Total seminars: 11 sessions (Week 2-15, no seminar on March 9, April 6, and April 13).
• You must earn at least 70 points to pass the course.
• Attendance: 7 points per seminar
• Asking one question during a seminar: 2 points

Absence Policy

• Absences may be excused with valid proof.
• Please notify the TA via NTU COOL or email if you cannot attend a seminar.

Student Speaker

• Students may volunteer to give a seminar talk.
• Student speakers are exempt from attending all the other seminars.
• Please email the Instructor if you are willing to give a talk by March 16.
• If a scheduled student talk is cancelled for any reason, the speaker will record and submit a video presentation.

Schedule (Tentative)

Week 2 (March 2)

• Speaker: 王新博 Hsin-Po Wang (National Taiwan University)

• Title: Group Testing in 6G: Downlink and Uplink

• Abstract: Group testing (GT) is a mathematical trick to identify a small number of targets from a large population using pooled tests, and it has become increasingly relevant in modern communications. For instance, for uplink, the challenge is to arrange devices who want to talk into frequency–time slots; for downlink, on the other hand, the challenge is to send messages to devices without Alice mistaking Bob’s message for her own. 6G will benefit from group testing tricks to support a massive number of devices with highly irregular activity.

  This talk applies GT to both downlink and uplink through a single design idea that we call cutting the plum pudding (CTPP). The analogy is simple: plums are randomly distributed in a pudding, and it is difficult to cut out exactly one plum. In our setting, the base station performs randomized cuts over device sets so that exactly one device is highlighted. Once the singleton is found, the task reduces to one-to-one communication, which is significantly easier to handle reliably.

• Bio: Hsin-Po Wang is an Assistant Professor at National Taiwan University (EE + GICE). He received a BSc in Mathematics from NTU and a PhD in Mathematics from UIUC, and he has worked at UC San Diego, UC Berkeley, and the Simons Institute before joining NTU.

  Hsin-Po Wang is interested in applying math tools such as algebra, combinatorics, calculus, and probability theory to information theory and coding. Particular topics he has worked on include polar codes (for wireless communication), group testing (this talk), regenerating codes (for cloud storage), distributed matrix multiplication (for cloud computation), DNA digital data storage (for very long-term storage), differential privacy (for privacy), exact distribution shaping (for randomized algorithms and ML), and pessimistic cardinality estimation (for database optimization).

Week 3 (March 9): no seminar, instructor attending Dagstuhl Seminar

Week 4 (March 16)

• Speaker: 黃上恩 Shang-En Huang (National Taiwan University)

• Title: Edge-Disjoint Paths in Expanders

• Abstract: We design efficient deterministic algorithms for finding short edge-disjoint paths in expanders. Before our work, deterministic polynomial-time algorithms were known only for expanders with constant conductance and were significantly slower. To obtain our result, we give an almost-linear time algorithm for hypergraph perfect matching under generalizations of Hall-type conditions (Haxell 1995), a powerful framework with applications in various settings, which until now has only admitted large polynomial-time algorithms (Annamalai 2018). In this talk, I will first explain the results, and give more details about the proofs of Haxell conditions and the analysis of the algorithm.

• Bio: Shang-En Huang is currently an assistant professor at NTU CSIE. He graduated from University of Michigan in 2022 and was a postdoc at Boston College. His research interests include dynamic graph data structures, graph sparsification, and distributed graph algorithms.

Week 5 (March 23): no seminar, instructor supporting EE1006

Week 6 (March 30): student talk

• Speaker: Hsiang-Shian Fan (National Taiwan University)

• Title: Approximate Minimum Selection with Persistent Random Comparison Faults

• Abstract: We study the problem of finding an element of rank $O(\log n)$ among $n$ elements with high probability in the presence of persistent random comparison faults. In this fault model, the outcome of comparing two elements is wrong with a constant error probability $p < 1/2$, comparing the same pair of elements multiple times always gets the same outcome, and the comparison faults between different pairs of elements are independent. In this paper, we develop an $O(n)$-time algorithm that, for any constant error probability $p < 1/2$, finds an element of rank $O(\log n)$ among $n$ elements with high probability. We also show that no algorithm can find an element of rank $o(\log n)$ with high probability. Moreover, the $O(n)$ time is optimal since $\Omega(n)$ sample elements are necessary to guarantee the existence of an element of rank $O(\log n)$ with high probability.

Week 7 (April 6): no seminar, spring break

Week 8 (April 13): no seminar, week of midterm

Week 9 (April 20): student talk

• Speaker: Jiahang Pang (National Taiwan University)

• Title: Innovative design of PVT collector for floating thermal contact

• Abstract: Conventional photovoltaic–thermal (PVT) collectors often face challenges such as high packaging costs, reliability issues arising from thermal stress, and the need for expensive thermal interface materials (TIMs). These challenges have limited their widespread adoption, despite the significant potential of PVT technology to improve overall energy efficiency. This paper presents an innovative design for a solar PVT collector featuring a plastic heat collector (PHC) in a floating thermal contact configuration. In this design, the PHC is coupled to the PV module using lubricating grease as the TIM, ensuring efficient thermal coupling while accommodating thermal expansion and minimizing mechanical stress. A thermal and electrical performance model was developed and validated to evaluate the proposed design, and long-term simulations were conducted for a PVT system consisting of a 9.6 kWp PV array and a 2,400 L thermal storage tank. The results show that employing low-cost lubricating grease (KTim = 0.41 W/mK) as the TIM enables the system to recover approximately 24–29 % of the incident solar irradiance as thermal energy while yielding an overall primary energy efficiency of roughly 71 %. Compared with a PV-only system, the proposed design reduces primary energy demand by about 54 %, while the PVT electrical output increases by approximately 6 % due to the cooling effect. Under representative tariffs (e.g., electricity \$0.12/kWh; natural gas \$0.06/kWh-th), the proposed PVT system shortens the whole-system payback by 3–5 years relative to PV-only.

Week 10 (April 27)

• Speaker: 俞韋亘 Wei-Hsuan Yu (National Central University)

• Title: The classical problems in discrete geometry

• Abstract: In this talk, I will introduce the classical research problems in discrete geometry such as the sphere packing problems and kissing number problems. Maryna Viazovska solved the sphere packing problems in dimensions 8 and 24, and then she got the fields medal in 2022. I will also mention the notion of spherical designs and few distance sets. Then, I will mention my results about the maximum size of spherical two-distance sets in Euclidean space and Hamming space.

Week 11 (May 4): student talk

• Speaker: Reza Jafari (National Taiwan University)

• Title: Embedding Quantum Program Verification into Dafny

• Abstract: Qafny, a compiler that embeds quantum program verification into the classical verifier Dafny. Qafny automatically translates high-level Qafny programs and their specifications into Dafny code that can be verified using SMT-based automated reasoning.

• Bio: I am a PhD student under Prof. Shih-Han Hung and Prof. Yu-Fang Chen working on quantum computing-related subjects. I also collaborate with Iowa State University. I developed a quantum network simulator for Nokia Bell Labs, and I am also interested in quantum networks and key distribution.

Week 12 (May 11)

• Speaker: 王喬萱 Chiao-Hsuan Wang (National Taiwan University)

• Title: Quantum Engineering for Analog Quantum Computation

• Abstract: Eigenvalue estimation is a central problem for demonstrating quantum advantage, yet its implementation on digital quantum computers is limited by circuit depth and operational overhead. To address these challenges, we present an analog quantum phase estimation (aQPE) protocol that extracts the eigenenergies of a target Hamiltonian through continuous time evolution and single-cavity readout. By encoding eigenvalue information as conditional cavity phase-space rotations, the scheme avoids deep quantum circuits and entangling gates, while enabling measurement with established cavity tomography. We further illustrate the feasibility of this approach with an implementation of aQPE for the XY model, whose ground-state energy problem is QMA-complete, within a physical architecture compatible with existing circuit quantum electrodynamics technology. These results demonstrate a resource-efficient and scalable framework for realizing quantum phase estimation on near-term platforms.

• Bio: Chiao-Hsuan Wang is an Assistant Professor in the Department of Physics at National Taiwan University. She received her Ph.D. in Physics from the University of Maryland, College Park, USA, in 2018, followed by postdoctoral research at Yale University and the University of Chicago from 2018 to 2022. She has held the NTU Eminence Scholar Fellowship since 2022 and has been a Center Scientist at the National Center for Theoretical Sciences, Physics Division, since 2023. Her research focuses on quantum information science, particularly fault-tolerant quantum computation and quantum communication, with broader interests in quantum optics and light–matter interactions.

Week 13 (May 18)

• Speaker: 吳廸融 Ti-Rong Wu (Academia Sinica)

• Title: Toward Efficient Planning in Deep Reinforcement Learning

• Abstract: Deep Reinforcement Learning (DRL) has achieved significant advancements in various domains, such as game-playing, robotics, and natural language processing. Among these fields, game-playing serve as an important testing ground for DRL algorithms due to their controllable and accessible environments that stand in contrast to the complex real-world problems. In this talk, I will present recent planning methods for game AI, focusing on how to improve the efficiency and scalability of AlphaZero/MuZero-style algorithms. First, I will introduce OptionZero, which incorporates learned options into planning, allowing an agent to discover temporally extended actions through self-play and search deeper under limited simulation budgets. I will then discuss Regret-Guided Search Control (RGSC), which improves AlphaZero training by identifying high-regret states and reusing them as informative starting positions for self-play. Together, these works show that better planning mechanisms can lead to more efficient and robust deep reinforcement learning.

• Bio: Ti-Rong Wu is currently an Assistant Research Fellow/Professor at the Institute of Information Science, Academia Sinica, Taiwan. He received his Ph.D. degree in Computer Science from National Chiao Tung University in 2020 and joined Academia Sinica in 2022. He led a team in developing a computer Go program, named CGI, which has won multiple gold medals in Computer Olympiad, and won second place in the first World AI Open in 2017. His current research interests include reinforcement learning, planning, and computer games.

Week 14 (May 25)

• Speaker: 葉宗泰 Tsung-Tai Yeh (National Yang Ming Chiao Tung University)

• Title: Energy-Efficient LUT-based GEMM Accelerator with Hardware-Aware KV Cache Quantization

• Abstract: The energy efficiency of Large Language Models (LLMs) model inference is becoming critical for edge AI applications. Recent quantization produces low-bit LLMs (1.58-bit weights) while minimizing the accuracy loss during model inference. However, the KV cache significantly increases model capacity, and the LLM decoding stage begins to dominate overall inference time as the context length grows. This talk will present our research on the energy-efficient GEMM accelerator for low-bit and mixed-precision LLM models. I will introduce our proposed Omni-LUT GEMM hardware accelerator architecture and a novel hardware-aware KV-cache quantization method. Our proposed hardware-software co-design architectural support enables efficient mixed-precision computations in the prefill and decode stages. Evaluations show that our proposed Omni-LUT accelerator significantly reduces model inference latency and energy consumption compared to state-of-the-art AI accelerators.

• Bio: Dr. Tsung Tai Yeh received his Ph.D degree from the Department of Electrical and Computer Engineering at Purdue University in 2020. He is currently an Associate Professor in the Computer Science (CS) Department at National Yang Ming Chiao Tung University (NYCU), Taiwan. Dr. Yeh has also been recognized with several awards. He received a Fellowship of the HEA from Advance HE in 2023, as well as the Excellent Mentor Award from NYCU in 2021 and 2025. He won the 2025 Qualcomm Innovation Fellowship. He is also a recipient of the 2030 Cross-Generation Young Scholars Program (NSTC), Taiwan, 2025. His research interests span computer architecture, computer systems, and programming languages, with a primary focus on the GPU, Domain-Specific Accelerators, and AI compiler systems. His compiler research was nominated for the Best Paper Award at the PPoPP conference, and his work was published in multiple top-ranking conference proceedings (ISCA, ASPLOS, HPCA, PPoPP, ICRA, NeurIPS).

Week 15 (June 1): student talk

• Speaker: Jen-Kuang Liu (National Taiwan University)

• Title: Scaling KV-stores with the Explosive Growth of Memory Capacity via Schedulable Hash Table Expansion

• Abstract: The rapid evolution of memory technologies enables hosts to deploy high-capacity memory or dynamically expand capacity via advanced interconnects. Leveraging this memory abundance, hash-based key-value stores facilitate high-speed, large-scale indexing without secondary storage access. However, high-volume insertions incur substantial rehash overheads due to extensive item migrations. Although asynchronous rehashing is widely adopted to decouple migration from the critical path, we identify a throughput mismatch between concurrent insertions and background migrations. This disparity leads to excessive traversal depths, ultimately triggering severe memory stalls and page faults.

  To address the challenges posed by explosive data growth, this work proposes a scheduling mechanism that efficiently amortizes rehash costs. We design a hybrid hash table composed of level-adjustable segments to adapt the amortization ratio, treating each segment as a scheduling entity for regional rehash tasks. Informed by potential page faults and memory stalls, a priority-based scheduler dispatches rehash tasks to heterogeneous threads, staggering asynchronous rehash execution over extended intervals to ensure index stability. By harvesting memory stall cycles, the insertion-migration throughput mismatch is bridged without compromising index efficiency. To validate our design, we implemented ExpHash, a NUMA-aware, lock-free, in-memory key-value store. ExpHash outperforms state-of-the-art systems by 1.6× in throughput under billion-scale, insert-intensive workloads, while delivering robust stability and consistently low latencies.

• Bio: Jen-Kuang Liu is a Ph.D. student in the Graduate Institute of Electrical Engineering at National Taiwan University (NTU). His research focuses on storage systems and operating systems.

Week 16 (June 8): no seminar, week of final exam