Bulletin of the American Physical Society
APS March Meeting 2017
Volume 62, Number 4
Monday–Friday, March 13–17, 2017; New Orleans, Louisiana
Session X49: Robot Scientists and Machine Learning for Automated Modeling and Control of Complex SystemsInvited Undergraduate
|
Hide Abstracts |
Sponsoring Units: DBIO GSNP Chair: John Wikswo, Vanderbilt University Room: 396 |
Friday, March 17, 2017 8:00AM - 8:36AM |
X49.00001: The Adam and Eve Robot Scientists for the Automated Discovery of Scientific Knowledge Invited Speaker: Ross King A Robot Scientist is a physically implemented robotic system that applies techniques from artificial intelligence to execute cycles of automated scientific experimentation. A Robot Scientist can automatically execute cycles of hypothesis formation, selection of efficient experiments to discriminate between hypotheses, execution of experiments using laboratory automation equipment, and analysis of results. The motivation for developing Robot Scientists is to better understand science, and to make scientific research more efficient. The Robot Scientist `Adam' was the first machine to autonomously discover scientific knowledge: both form and experimentally confirm novel hypotheses. Adam worked in the domain of yeast functional genomics. The Robot Scientist `Eve' was originally developed to automate early-stage drug development, with specific application to neglected tropical disease such as malaria, African sleeping sickness, etc. We are now adapting Eve to work with on cancer. We are also teaching Eve to autonomously extract information from the scientific literature. [Preview Abstract] |
Friday, March 17, 2017 8:36AM - 9:12AM |
X49.00002: Automated inference of biological and physical models Invited Speaker: Hod Lipson |
Friday, March 17, 2017 9:12AM - 9:48AM |
X49.00003: Automated adaptive inference of phenomenological dynamical models Invited Speaker: Bryan Daniels Understanding the dynamics of biochemical systems can seem impossibly complicated at the microscopic level: detailed properties of every molecular species, including those that have not yet been discovered, could be important for producing macroscopic behavior. The profusion of data in this area has raised the hope that microscopic dynamics might be recovered in an automated search over possible models, yet the combinatorial growth of this space has limited these techniques to systems that contain only a few interacting species. We take a different approach inspired by coarse-grained, phenomenological models in physics. Akin to a Taylor series producing Hooke's Law, forgoing microscopic accuracy allows us to constrain the search over dynamical models to a single dimension. This makes it feasible to infer dynamics with very limited data, including cases in which important dynamical variables are unobserved. We name our method \emph{Sir Isaac} after its ability to infer the dynamical structure of the law of gravitation given simulated planetary motion data. Applying the method to output from a microscopically complicated but macroscopically simple biological signaling model, it is able to adapt the level of detail to the amount of available data. Finally, using nematode behavioral time series data, the method discovers an effective switch between behavioral attractors after the application of a painful stimulus. [Preview Abstract] |
Friday, March 17, 2017 9:48AM - 10:24AM |
X49.00004: Discovering governing equations from data by sparse identification of nonlinear dynamics Invited Speaker: Steven Brunton The ability to discover physical laws and governing equations from data is one of humankind’s greatest intellectual achievements. A quantitative understanding of dynamic constraints and balances in nature has facilitated rapid development of knowledge and enabled advanced technology, including aircraft, combustion engines, satellites, and electrical power. There are many more critical data-driven problems, such as understanding cognition from neural recordings, inferring patterns in climate, determining stability of financial markets, predicting and suppressing the spread of disease, and controlling turbulence for greener transportation and energy. With abundant data and elusive laws, data-driven discovery of dynamics will continue to play an increasingly important role in these efforts.\\ This work develops a general framework to discover the governing equations underlying a dynamical system simply from data measurements, leveraging advances in sparsity-promoting techniques and machine learning. The resulting models are parsimonious, balancing model complexity with descriptive ability while avoiding overfitting. The only assumption about the structure of the model is that there are only a few important terms that govern the dynamics, so that the equations are sparse in the space of possible functions. This perspective, combining dynamical systems with machine learning and sparse sensing, is explored with the overarching goal of real-time closed-loop feedback control of complex systems.\\ This is joint work with Joshua L. Proctor and J. Nathan Kutz.\\ Video Abstract: https://www.youtube.com/watch?v=gSCa78TIldg [Preview Abstract] |
Friday, March 17, 2017 10:24AM - 11:00AM |
X49.00005: Thinking in machines, not statistics Invited Speaker: Sarah Marzen When asked to summarize a long string of data, we can either model the trajectory distribution directly or infer machines that could have likely produced the observed trajectory. I will argue that thinking in terms of machines, rather than in terms of trajectory distributions, can lead to improved inference algorithms and more accurate plug-in estimators of various information-theoretic quantities. I will focus on the predictive information bottleneck as an illustrative example. [Preview Abstract] |
Follow Us |
Engage
Become an APS Member |
My APS
Renew Membership |
Information for |
About APSThe American Physical Society (APS) is a non-profit membership organization working to advance the knowledge of physics. |
© 2024 American Physical Society
| All rights reserved | Terms of Use
| Contact Us
Headquarters
1 Physics Ellipse, College Park, MD 20740-3844
(301) 209-3200
Editorial Office
100 Motor Pkwy, Suite 110, Hauppauge, NY 11788
(631) 591-4000
Office of Public Affairs
529 14th St NW, Suite 1050, Washington, D.C. 20045-2001
(202) 662-8700