AD1024 (real name: Mike He) is a Junior student at Paul G. Allen School of Computer Science & Engineering. His research interests include Program Synthesis, Programming Language Theory (specifically Type Theory related), Formal Verification and Compiler Designs. He is a member of SAMPL and PLSE Research Group. He loves classical music, anime and Japanese popculture.
-- Email (λx. λy. λz. [email protected]) cs washington.edu dh63
$^*$: Listed publication
3LA / Marlowe: A Verified Compilation Flow for Deep Learning Accelerators (Jun. 2020 $\Rightarrow$ Now)
$^*$Dynamic Tensor Rematerialization: A Heuristic-based Gradient Checkpointing Algorithm that supports automatic checkpointing on arbitrary Deep Learning Models (Nov. 2019 $\Rightarrow$ Now)
Relay: A High-Level Intermediate Representation for Deep Learning (Oct. 2019 $\Rightarrow$ 2020)
UW Solar Data: Data Monitoring Dashboard for Solar Panels (Jan. 2019 $\Rightarrow$ Jun. 2019)
Kirisame, M.*, Lyubomirsky, S.*, Haan, A.*, Brennan, J., He, M., Roesch, J., Chen, T., Tatlock, Z. (2020, June 18). Dynamic Tensor Rematerialization. ICLR 2021. September 19, 2020. https://arxiv.org/abs/2006.09616 (*: Equal contribution)
Zachary Tatlock, Sharad Malik, Aarti Gupta, Gu-Yeon Wei, Steven Lyubomirsky, Bo-Yuan Huang, Thierry Tambe, Mike He, Yi Li, and Gus Smith. 2021. From DSLs to Accelerator-rich Platform Im-plementations: Addressing the Mapping Gap. In LATTE ’21: Work-shop on Languages, Tools, and Techniques for Accelerator Design,April 15, 2021, Virtual Event. ACM, New York, NY, USA, 3 pages.
Note: bold stands for highly recommended courses.
I love classical music and enjoy playing the violin. I’ve been playing the violin for 0x11 years.
Some recordings are available here: AD1024 @ Bilibili