Acceptance rate: 22.5% (132/586)
Acceptance rate: 17.4% (31/178)
Best Paper Award (1/178)
Acceptance rate: 23%
Acceptance rate: ~24%
Acceptance rate: 23%
Acceptance rate: 17.6% (70/398)
Acceptance rate: 21.9%(28/128)
Best Paper Award
Acceptance rate: 24.8%(202/815)
Acceptance rate: 24.3%(168/691)
The First Prize Paper Award
IEEE PhD Thesis Award sponsored by IEEE Taipei Section. (IEEE 2021 碩博士論文獎)
One of the Finalist of Open Innovation Contest for AXDIMM Technology hosted by Samsung.
The 1st Prize of PhD Thesis Award sponsored by Lam Research. (科林論文獎博士論文頭等獎)
Student Travel Grants sponsored by Embedded Systems Week (ESWEEK).
Outstanding Students Conference Travel Grants sponsored by Foundation for the Advancement of Outstanding Scholarship (FAOS). (傑出人才發展基金會)
The International Conference Scholarship For Young Researchers sponsored by Academia Sinica. (中央研究院年輕學者出國報告)
Seminar Talk Speaker of NTHU ISA, Hsinchu, 2022.
Invited Talk Speaker of Meta, Boston, 2022.
Tutorial Speaker of The 24th Design, Automation and Test in Europe Conference (DATE), 2021.
Student Research Forum Speaker of The 26th Asia and South Pacific Design Automation Conference(ASP-DAC), 2021.
Invited Talk Speaker of The 8th IEEE Non-Volatile Memory Systems and Applications Symposium (NVMSA), 2019.
Researchers from Meta (or Facebook previously) point out that data preprocessing is becoming a critical performance bottleneck for training their recommendation systems. We observed that, one of the reasons of the bottleneck is that unused training data may still be read and filtered out during data preprocessing. Besides, these unused data movements is because of the access behavior gap between recommendation systems and SSDs. To avoid these unused data movements, We proposed a joint management middleware to bridge the access behavior gap and periodically re-organize the training data inside SSDs. With using our middleware, systems can save 24%-47% of the overall read time compared with the LSM-based strategy, which is now currently applied by Meta and Baidu.
We come up with a proposal targeting on integrating the GNN-based recommendation system with AxDIMM. We are now working on the architecture analysis of AxDIMM and the behavior analysis of GNN-based recommendation system.
Conduct researches, experiments, or implementations related to computer systems such as storage systems, memory systems, embedded systems, computer architecture, energy-efficient designs, multi-core/many-core systems, and neuromorphic computing.
This project aims to provide a private cloud storage for users. Our team designs a distributed storage system, SSBox, with high accessibility and reliability. We provide PaaS layer services for programmers to access our SSBox by RESTful API. In addition, SSBox could endure hundred of thousand of users to access simultaneously.
This project aims to provide Virtual Desktops to cost down the hardware price for schools. We apply a real-time virtual desktop based on OpenStack and Docker. Users just need a browser and stable Internet so that users can access different Operating System. We also design a client side by OpenStack APIs and the design makes users to create virtual desktops eaiser.