Use this URL to cite or link to this record in EThOS:
Title: Novel algorithms for the understanding of the chemical cosmos
Author: Makrymallis, A.
ISNI:       0000 0004 5359 3284
Awarding Body: University College London (University of London)
Current Institution: University College London (University of London)
Date of Award: 2015
Availability of Full Text:
Access from EThOS:
Full text unavailable from EThOS. Please try the link below.
Access from Institution:
Molecular data from the interstellar medium (ISM) contain information that holds the key to understanding our chemically controlled cosmos and to unlocking the secrets of our universe. Observational data, as well as synthetic data from chemical codes, provide a cornucopia of digital information that conceals knowledge of the ISM. Astrochemistry studies the chemical interactions in the ISM and translates this information into knowledge of the physical characteristics of the ISM. As larger datasets and more complex models are being employed in astrochemistry, the need for intelligent data mining algorithms wil increase. Machine learning algorithms provide novel methods for human-driven analysis of astrochemical data by augmenting scientific intelligence. The aim of this thesis is to introduce machine learning methods for solving typical astrochemical problems. The main application focus will be the physical parameter profile of dark molecular clouds. Time-dependent chemical codes are typically used as a tool to interpret observations, but their potential to explore a large physical and chemical parameter space is often ne- glected due to the computational complexity or the complexity of the parameter space. We will present clustering analysis methods, using traditional and probabilistic hierar- chical clustering, for the efficient discovery of structure and patterns in vast parameter spaces generated solely from an astrochemical code. Moreover, we will demonstrate how Bayesian methods in conjunction with Markov Chain Monte Carlo sampling algorithms can efficiently solve nonlinear inverse problems for the probabilistic estimation of chemical and physical parameters of dark molecular clouds. The computational cost of sampling algorithms can be preventive for a full Bayesian approach in some cases, hence we will also present how artificial neural networks can accelerate the inference process without much loss of accuracy. Finally, we will demonstrate how the Bayesian approach and smart sampling techniques can tackle uncertainty about surface reactions and rate coefficients, even with vague and not very informative observational constraints, and assist laboratory astrochemists by guiding experimental techniques probabilistically.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available