- Library Home /
- Search Collections /
- Open Collections /
- Browse Collections /
- UBC Theses and Dissertations /
- New approaches at crossroads between molecular physics...
Open Collections
UBC Theses and Dissertations
UBC Theses and Dissertations
New approaches at crossroads between molecular physics and machine learning Asnaashari, Kasra
Abstract
This thesis explores the interplay of machine learning and molecular physics, demonstrating how developments in one field lead to advances in the other. We show this using two examples. First, we illustrate how improvements made to a machine learning model of molecular PES and FF can significantly increase its generalization accuracy. Gradient domain machine learning (GDML) models produce accurate results for high-dimensional molecular systems with a small number of ab initio calculations. The present work extends GDML to models with composite kernels built to maximize inference from a small number of molecular geometries. We illustrate that GDML models can be improved by increasing the complexity of underlying kernels through a greedy search algorithm using Bayesian information criterion as the model selection metric. We show that this requires including anisotropy into kernel functions and produces models with significantly smaller generalization errors. The results are presented for ethanol, uracil, malonaldehyde and aspirin. For aspirin, the model with composite kernels trained by forces at 1000 randomly sampled molecular geometries produces a global 57-dimensional PES with the mean absolute error 0.177 kcal/mol (61.9 cm⁻¹) and FFs with the mean absolute error 0.457 kcal/mol Å⁻¹. Second, we propose a procedure to perform quantum computation in the form of quantum annealing using a crossing between Zeeman states of different parity in the rotational and fine structure of open-shell molecules in a Σ electronic state. These crossings become avoided in the presence of an electric field. We propose an algorithm that encodes Ising models into qubits defined by pairs of ²Σ molecules sharing an excitation near these avoided crossings. This can be used to realize a transverse field Ising model tunable by an external electric or magnetic field, suitable for quantum annealing applications. We perform dynamical calculations for several examples with 1D and 2D connectivities. Our results demonstrate that the probability of obtaining valid annealing solutions is high and can be optimized by varying the annealing times.
Item Metadata
Title |
New approaches at crossroads between molecular physics and machine learning
|
Creator | |
Supervisor | |
Publisher |
University of British Columbia
|
Date Issued |
2022
|
Description |
This thesis explores the interplay of machine learning and molecular physics, demonstrating how developments in one field lead to advances in the other. We show this using two examples. First, we illustrate how improvements made to a machine learning model of molecular PES and FF can significantly increase its generalization accuracy. Gradient domain machine learning (GDML) models produce accurate results for high-dimensional molecular systems with a small number of ab initio calculations. The present work extends GDML to models with composite kernels built to maximize inference from a small number of molecular geometries. We illustrate that GDML models can be improved by increasing the complexity of underlying kernels through a greedy search algorithm using Bayesian information criterion as the model selection metric. We show that this requires including anisotropy into kernel functions and produces models with significantly smaller generalization errors. The results are presented for ethanol, uracil, malonaldehyde and aspirin. For aspirin, the model with composite kernels trained by forces at 1000 randomly sampled molecular geometries produces a global 57-dimensional PES with the mean absolute error 0.177 kcal/mol (61.9 cm⁻¹) and FFs with the mean absolute error 0.457 kcal/mol Å⁻¹.
Second, we propose a procedure to perform quantum computation in the form of quantum annealing using a crossing between Zeeman states of different parity in the rotational and fine structure of open-shell molecules in a Σ electronic state. These crossings become avoided in the presence of an electric field. We propose an algorithm that encodes Ising models into qubits defined by pairs of ²Σ molecules sharing an excitation near these avoided crossings. This can be used to realize a transverse field Ising model tunable by an external electric or magnetic field, suitable for quantum annealing applications. We perform dynamical calculations for several examples with 1D and 2D connectivities. Our results demonstrate that the probability of obtaining valid annealing solutions is high and can be optimized by varying the annealing times.
|
Genre | |
Type | |
Language |
eng
|
Date Available |
2022-08-26
|
Provider |
Vancouver : University of British Columbia Library
|
Rights |
Attribution-NonCommercial-NoDerivatives 4.0 International
|
DOI |
10.14288/1.0417579
|
URI | |
Degree | |
Program | |
Affiliation | |
Degree Grantor |
University of British Columbia
|
Graduation Date |
2022-11
|
Campus | |
Scholarly Level |
Graduate
|
Rights URI | |
Aggregated Source Repository |
DSpace
|
Item Media
Item Citations and Data
Rights
Attribution-NonCommercial-NoDerivatives 4.0 International