Speaker: Professor Lulu Qian, California Institute of Technology
Title: "DNA neural networks that learn"
Abstract: Learning allows living systems to be born simple but develop into endless diversity and richness; it also accelerates evolution by altering the shape of the fitness landscape -- known as the Baldwin effect. It has been hypothesized for about 30 years that neural computation occurs not only at the multicellular level in the brain, but also at the molecular level within individual cells. It was further proposed that learning rules observed in the brain could in principle be carried out by molecular circuits, for example in genetic regulatory networks in vivo and in DNA strand-displacement circuits in vitro. Experimental demonstration of neural computation in DNA circuits provided a proof of concept that rudimentary brain-like behavior can exist in test tube chemistry, and opened the possibility of embedding learning within engineered molecular systems so that they can become “smarter” over their lifetimes.
In this talk, I will show how a collection of DNA molecules can be programmed to perform supervised learning, where the system is exposed to examples of what it may encounter and examples of the desired response; this information is used to develop memories and improve the molecular system’s capability for handling similar situations in the future. I will also discuss key challenges toward unsupervised learning, where no teacher is present, and the molecular system must spontaneously learn from an unknown environment.
Biography: Dr. Lulu Qian is a Professor of Bioengineering and an affiliated faculty of Computer Science and Computation & Neural Systems at the California Institute of Technology. The primary focus of her lab is to advance the theory and practice of engineering molecular systems with intelligent behaviors. She takes inspiration from fundamental principles in biology and conceptual frameworks in computer science to develop systematic approaches for the construction of artificial molecular machines. She is a recipient of the Burroughs Wellcome Fund Career Award at the Scientific Interface, the National Science Foundation Faculty Early Career Development Award, the Rozenberg Tulip Award in DNA Computing, the Foresight Institute Feynman Prize in Nanotechnology, the Schmidt Science Polymath Award, and the Feynman Prize for Excellence in Teaching.