Tianqin Li
The ability for the human mind to logically reason has been a key distinguishing feature to our species success. Our capacity to distill our observations of the world into discrete composable concepts enables us to extract critical information from an ocean of noise and efficiently communicate our ideas. Using these composable prototypical concepts, we can further extend previously learned knowledge to abstract unseen observations and do more hierarchical planning and out-of-domain reasoning.
These attributes of the human mind stand in sharp contrast to modern artificial neural networks, which represent input as entangled and unstructured vectors, not robust to domain shift. This contrast motivates me to build smart machines that can (1) perform efficient and robust reasoning from raw sensory information; (2) learn with limited supervision.
In addition, I'm also interested in various computer vision tasks, including 3D synthesis, dynamic point clound sequence enhancement, video prediction and interpolation, etc. Before joining CMU and doing AI research, I studied bioinformatics and developed softwares for synthetic biologists to automate the process of artificially engineering life.
Collected Quotes
“Shoot for the moon. Even if you miss, you'll land among the stars.”
“Make everything as simple as possible, but not simpler.”
“A problem well stated is a problem half-solved.”
"Major discoveries are almost always preceded by bewildering, complex observations ... I always believed that the neocortex appeared complicated largely because we didn't understand it, and that it would appear relatively simple in hindsight. Once we knew the solution, we would look back and say, 'Oh, of course, why didn't we think of that?' When our research stalled or when I was told that the brain is too complicated to understand, I would imagine a future where brain theory was part of every high school curriculum. This kept me motivated."
News
[Presentation] I presented the Top-K for Shape Bias in NeurIPS 2023 in oral session 2B at Ballroom Hall A - C in New Orleans Conventional Center (Dec 12 2023)!
[Publication] The work Top-K for Shape Bias is accepted by NeurIPS 2023 as an Oral Paper!
[Ph.D. Start] Got accepted in Carnegie Mellon University Computer Science Department for Ph.D. study starting Fall 2022.
[Publication] The work MoCA is accepted by ICLR 2022.
[Publication] The work Cl-InfoNCE is accepted by ICLR 2022.
[Publication] The work CCLK is accepted by ICLR 2022.
[Publication] The work TPU-GAN is accepted by ICLR 2022.
[Publication] The work SurfGen is accepted by ICCV 2021.
Projects
Emergence of Shape Bias in Convolotional Neural Networks through Activation Sparsity
Tianqin Li, Ziqi Wen, Yangfan Li, Tai Sing Lee.
Oral (session 2B) presentation at Ballroom Hall A - C in New Orleans Conventional Center (Dec 12 2023)!
Accepted by the 37th Neural Information Processing System (NeurIPS 2023 - Oral - selective 1%)
Paper | Project Page | Code
