Statistical learning in quantum chemistry
Abstract/Contents
- Abstract
- This dissertation presents generalized paradigms for combining the most powerful aspects of quantum chemistry and machine learning to provide relevant, accurate and rapid predictions of objective quantities in chemical and materials systems. The focus of the approaches herein is on eliminating the primary areas of deficiency in quantum chemistry, which limit its ability to independently direct applied, experimental research. Using statistical techniques such as Gaussian Process regression, nonlinear optimization, ensemble learning and Bayesian global optimization, quantum chemical modeling errors arising from model inadequacy or inefficiency may be quantified and mitigated in a computational and data efficient manner. Further, incorporating such statistical techniques directly into traditional quantum chemistry models allows for transformative modeling capabilities in the form of predictive uncertainty estimates and systematic, incremental model improvement. The applications addressed here broadly fall into the category of materials design, with specific attention given to high throughput screening of OPV materials, protein folding and photo-excitation dynamics in light-harvesting systems. The formative and ultimate goal of this discourse is to shift the paradigm in quantum chemistry from a focus on the ontological (i.e. descriptive) capabilities of modeling, to the pragmatic (i.e. predictive). In practice, a combination of both approaches allows for computation to interpret and direct experimental processes in real time. As experimentation continues to progress toward rapid, automated chemical and materials prototyping, such computational methods will provide an invaluable counterpart in the form of efficient, automated directives toward materials optimization and discovery.
Description
Type of resource | text |
---|---|
Form | electronic; electronic resource; remote |
Extent | 1 online resource. |
Publication date | 2016 |
Issuance | monographic |
Language | English |
Creators/Contributors
Associated with | Sisto, Aaron |
---|---|
Associated with | Stanford University, Department of Materials Science and Engineering. |
Primary advisor | Martinez, Todd J. (Todd Joseph), 1968- |
Primary advisor | Reed, Evan J |
Thesis advisor | Martinez, Todd J. (Todd Joseph), 1968- |
Thesis advisor | Reed, Evan J |
Thesis advisor | Markland, Thomas E |
Thesis advisor | Prinz, F. B |
Advisor | Markland, Thomas E |
Advisor | Prinz, F. B |
Subjects
Genre | Theses |
---|
Bibliographic information
Statement of responsibility | Aaron Sisto. |
---|---|
Note | Submitted to the Department of Materials Science and Engineering. |
Thesis | Thesis (Ph.D.)--Stanford University, 2016. |
Location | electronic resource |
Access conditions
- Copyright
- © 2016 by Aaron Sisto
Also listed in
Loading usage metrics...