The Use of Machine Learning and AI to Improve Computational Performance in Large-Scale Optimization and Time Series Applications
Guardat en:
| Publicat a: | ProQuest Dissertations and Theses (2024) |
|---|---|
| Autor principal: | |
| Publicat: |
ProQuest Dissertations & Theses
|
| Matèries: | |
| Accés en línia: | Citation/Abstract Full Text - PDF |
| Etiquetes: |
Sense etiquetes, Sigues el primer a etiquetar aquest registre!
|
| Resum: | The three essays in my dissertation proposal examine the use of machine learning and artificial intelligence (ML/AI) for performance improvement in large-scale combinatorial problems and time series forecasting. Success in the application of ML/AI to these two areas has been limited in contrast to the spectacular success of ML/AI in many other application areas.My first essay surveys recent research on the use of ML/AI techniques to improve computational performance in large-scale combinatorial optimization (CO) problems. The survey concludes that ML/AI approaches have promise but that a general approach to solving a broad class of CO problems has yet to be discovered.The second essay investigates the use of large language models (LLMs) to automatically set the parameters and determine the best configuration for commercial solvers, which may have over one hundred parameters controlling every aspect of algorithm performance. I develop prompts and a strategy for employing an LLM in this task. Promising results were obtained in experiments restricted to cutting-plane selection.My third essay explores uncertainty quantification when ML/AI techniques are applied to forecasting financial time-series data. I experiment with methods to evaluate aleatoric uncertainty due to the predictive model) and epistemic uncertainty (due to missing data.) I develop a Bayesian Neural Network procedure that captures the relation between the exogenous and endogenous variables. Imposing a prior distribution on the network parameters captures uncertainty due to parametrization via the posterior distribution which combines the likelihood and the prior. Deep neural models are used to decompose uncertainty into aleatoric and epistemically components. |
|---|---|
| ISBN: | 9798346879343 |
| Font: | ProQuest Dissertations & Theses Global |