Services

ICML 2022 reviewer

Presentation at Vector Institute research symposium "RePU is All You Need"

DEQ Poster presentation at ACML2021

Volunteer at CCF-ADL2021

Significant Bits

"Can Few-Shot Learners Generalize to Unseen Compositions of Novel Primitives?" coming soon Withdrawn from ICLR2022, under review at ICJAI 2022

One patent coming soon: PT2021CN043(KP211671)

Composability in Julia: Implementing Deep Equilibrium Models via Neural ODEs Qiyao Wei, Frank Schäfer, Avik Pal, Chris Rackauckas, Julialang, 2021 Blog / Code / News Taking advantage of composability in Julia, we reformulate DEQs using ideas from NeuralODE, leading to a flexible implementation that could easily be extended.

Open-source Contributions

	PettingZoo
	DiffEqFlux
	SparseDiffTools

Less Significant Bits

• I implemented distributed training in response to this github issue
• I implemented a web-scraping pipeline in Scrapy and Redis
• I implemented gradient descent from scratch in numpy here
• I made snowflake animations following Brownian motion here