Jacob Gardner
University of Pennsylvania, PA·Computer Science·Early-career
Many scientific discoveries can be framed as optimization problems. A chemist wants a molecule with the strongest antiviral activity; an engineer wants the most aerodynamic vehicle.
- Fields
Computer Science
Mathematics & Statistics
Engineering
- Topics
black-box optimization
Bayesian optimization
machine learning
- Grants
- 1 known grant, through Sep 2027
What they work on
- Develop black-box optimization methods that scale to high-dimensional, structured search problems such as molecule design, where each candidate must be tested experimentally.
- Build deep representation models that map discrete objects such as graphs, strings, point clouds and images into continuous latent spaces suited to optimization, using unsupervised and multi-task data.
- Create large-scale probabilistic machine learning methods that balance exploitation and exploration during optimization, applied to therapeutic design, Alzheimer's-related cognitive science and energy technology.
More about the research
Often the property can only be measured by experiment, so the search is "black-box": each test is costly and there is no formula to follow. Such searches get much harder when the objects are complex structures, such as molecules.
Why it matters
Scientists designing drugs, energy technologies and engineered systems often must run expensive experiments to test each candidate. Data-driven optimization methods could help them find promising molecules or designs with fewer experiments. This could speed work on new therapeutics and energy technologies.
Known grants
From NIH and NSF research grants. Labs often have other funding we can't see.
Closest labs
Pan Xu
Duke University, NC·Computer Science·Early-career
Develop distributionally robust reinforcement learning methods that use minimax optimization over a set of possible environment dynamics, so that policies trained in a simulator perform reliably in an unknown target domain, with finite-sample guarantees.
Chao Zhang
Georgia Institute of Technology, GA·Computer Science·Early-career
Develop deep predictive models for spatial networks that capture how they change over space and time while estimating the uncertainty of their own forecasts.
Ziv Scully
Cornell University, NY·Computer Science·Early-career
Develop Gittins index acquisition functions for exotic Bayesian optimization problems, so that searches over possible AI models find higher-quality models with less training time and energy.
The school
University of Pennsylvania: 90 labs on PI Pages, $2.2B a year on research, 750 research doctorates a year.
Reach out
A start for your email. Fill in the brackets.