Tai Sing Lee
Carnegie Mellon University, PA·Neuroscience
The human brain uses millions of times less energy than today's AI systems. Yet it is more flexible, partly because it learns concepts and how they relate to each other instead of memorizing huge datasets.
- Fields
Neuroscience
Computer Science
- Topics
visual cortex
recurrent circuits
neural plasticity
- Grants
- 1 known grant, through May 2028
What they work on
- Develop a machine learning framework that formalizes the rapid recurrent plasticity, or fast rewiring, of local circuits in the early visual cortex.
- Test whether recurrent circuits reshape neural activity space so that related concepts move closer together while irrelevant variation is compressed, using associative learning, denoising and pattern completion tasks.
- Combine computational modeling with neurophysiological experiments to evaluate how well these brain-inspired models predict results, and how useful they are in computer vision applications.
More about the research
Neurons in the early visual cortex, the brain's first stage of visual processing, mainly detect local features such as edges. They can also quickly form recurrent circuits, loops of connections among nearby neurons that let each neuron's response reflect the wider scene. Why the brain uses this rapid wiring is not well understood.
Why it matters
Understanding how visual cortex circuits encode relationships between concepts could guide the design of AI systems that are more energy-efficient and better at generalizing. These systems would be modeled on the brain's flexible intelligence. The models will be tested in real computer vision applications. The work also offers neuroscience a new framework for understanding recurrent circuits in the cortex.
Known grants
From NIH and NSF research grants. Labs often have other funding we can't see.
Closest labs
Michael Stryker
University of California-San Francisco, CA·Neuroscience
Record large populations of neurons with multi-electrode arrays in the retina and visual cortex of awake, behaving mice while they view naturalistic visual flow and simpler artificial patterns.
Christopher Kanan
University of Rochester, NY·Computer Science·Early-career
Develop brain-inspired continual learning algorithms that learn online and then run sleep-like consolidation periods, based on complementary learning systems theory, so deep neural networks gain new knowledge without forgetting old knowledge.
Adam Kohn
Albert Einstein College of Medicine, NY·Neuroscience
Test whether feedforward–feedback loops between visual areas carry out predictive coding, by analyzing population responses recorded at the same time in macaque V1/V2 and V1/V4 to targeted visual stimuli.
The school
Carnegie Mellon University: 28 labs on PI Pages, $505M a year on research, 346 research doctorates a year.
Reach out
A start for your email. Fill in the brackets.