Saudi Cultural Missions Theses & Dissertations
Permanent URI for this communityhttps://drepo.sdl.edu.sa/handle/20.500.14154/10
Browse
2 results
Search Results
Item Restricted Improving Model-fitting for Computational Models of Neural Function and Behaviour(King's College London, 2024) Ramis, Sarah; wise, TobyParameter estimation plays a critical role in computational neuroscience, enabling the design of models that can predict neural mechanisms under various conditions using statistical methods. However, accurate estimation becomes challenging when the likelihood function is intractable due to the model’s complexity and high dimensionality. With the advent of artificial intelligence, various machine learning algorithms have been developed to estimate the posterior distribution of observed data using simulation-based inference approaches. This study aims to elucidate the capabilities of these machine learning models compared to conventional methods to aid in model selection. Rescorla-Wagner simulated data is used to compare the performance of several techniques: Markov Chain Monte Carlo (MCMC) as a conventional method, Automatic Posterior Transformation (APT) using the Masked Auto-regressive Flow model, and APT enhanced with Gated Recurrent Units (APT-GRU) for data embedding, to estimate the parameters alpha and beta. The experiment involved 12 trials with varying parameter pairs and data volumes. The models were evaluated based on computational time, Mean Squared Error (MSE), Mean Absolute Error (MAE), a customized fit metric, Kullback-Leibler (KL) divergence, convergence, and log-likelihood. Results indicated that computational time increased across all models as dataset volume grew, with MCMC requiring significantly more time than the others. APT-GRU achieved lower MSE and MAE scores than the other models, indicating higher accuracy, followed by APT. Both APT-based models demonstrated stable convergence on large datasets, with APT-GRU showing enhanced convergence on smaller datasets. APT models also approached zero in KL divergence scores and positive log-likelihood values, signifying a good model fit. MCMC, while performing better in estimating alpha than beta, showed large negative KL divergence values, suggesting underfitting. Furthermore, MCMC exhibited lower performance stability across trials, with higher estimation variances. In summary, the findings highlight the strengths of APT and APT-GRU in parameter estimation within the Rescorla-Wagner framework. They suggest that the inclusion of GRU for data embedding improves both model robustness and estimation accuracy. These insights pave the way for future research into exploring different machine learning algorithms and model architectures for parameter estimations.12 0Item Restricted Hierarchical Bayesian Modelling of Spatio-Temporal Air Pollution Patterns in Athens, Greece(University of Sheffield, 2024) Aldawsari, Hilah; Johnson, Jill SAir pollution is one of the biggest health risks of our time, mainly due to industrial development worldwide. Persistent exposure to air pollution causes many severe diseases, including lung cancer. It also is responsible for millions of deaths annually worldwide. As a result, various researchers have been developing models to estimate air pollution levels throughout the past decades, aiming to better understand their dynamics. These models can also be used to forecast air pollution levels. This forecasted information is crucial for enabling relevant authorities to adopt precautionary measures and to alert the public in advance, especially when the levels are very harmful to human health. Up to date, modelling air pollution is a challenging task due to the various factors that affect its behaviour. One advanced field that enables the modelling of air pollution levels is spatio-temporal modelling. The main focus of this dissertation is to develop a spatio-temporal model to estimate parameters and forecast air pollution levels across different locations in Athens, Greece. Among various pollutants, this work focuses on exploring the ozone (O3) pollutant, along with three meteorological variables, to examine their influence on O3 levels. Two statistical approaches were developed for modelling and forecasting monthly concentrations of O3 in Athens, Greece. The first approach is based on time series modelling using seasonal autoregressive integrated moving average (SARIMA) and seasonal autoregressive integrated moving average with exogenous variables (SARIMAX) models. The second approach is based on spatio-temporal modelling using hierarchical Bayesian spatiotemporal Gaussian process (GP) models. This model has the ability to capture the complex dynamic of O3 levels, which vary based on several factors, including geographic location and meteorology. This model also has the ability to provide predictions for both space and time. Both SARIMA and SARIMAX models showed a reasonable fit of the data. The SARIMAX outperformed SARIMA models in both training and testing tests in most stations. The hierarchical Bayesian spatio-temporal GP model fits the data well and shows reasonable performance in both forecasting and prediction abilities, with a spatial decay parameter ϕ of 0.13 corresponding to an effective range of 23 km. All statistical data analysis in this dissertation was implemented using the statistical software R.21 0