High-dimensional and spatial data analysis by Bayesian inference
Lead Research Organisation:
Brunel University
Department Name: Computer Science
Abstract
Project summary: Whereas (1) standard regression models such as mean regression (ordinary least-squares regression) have been widely applied in many fields such as economics, genetics, ecology and environment and attracted huge amount of attention in literature, and (2) Bayesian inference is one of popular methods recent decades, there is relatively little work on Bayesian inference of regression models beyond mean for problems in those fields.
This project aims to: develop novel methods of Bayesian autoregressive beyond mean, propose new single-index models for high-dimensional data analysis, explore effective spatial regression beyond mean, compare different algorithms for Bayesian inference of regression models above.
This project aims to: develop novel methods of Bayesian autoregressive beyond mean, propose new single-index models for high-dimensional data analysis, explore effective spatial regression beyond mean, compare different algorithms for Bayesian inference of regression models above.
People |
ORCID iD |
| Keming Yu (Primary Supervisor) | |
| SANNA SOOMRO (Student) |
Studentship Projects
| Project Reference | Relationship | Related To | Start | End | Student Name |
|---|---|---|---|---|---|
| EP/R512990/1 | 30/09/2018 | 29/09/2023 | |||
| 2295266 | Studentship | EP/R512990/1 | 30/09/2019 | 30/05/2023 | SANNA SOOMRO |
| Title | Asymmetric Nonconvex Huber loss function |
| Description | The asymmetric nonconvex Huber loss function takes both robustness and assymetry simultaneously into account of location parameter estimation and distribution fitting. This results in a class of new asymmetric nonconvex Huber distribution. |
| Type Of Material | Computer model/algorithm |
| Year Produced | 2021 |
| Provided To Others? | No |
| Impact | Such a class of loss functions can lead to an elegant probability distribution, which is a normal scale-mixture and then has wide applications in Bayesian inference non-normal noise with Gibbs sampling. EM algorithm is developed based parameter estimation to solve the proposed probability distribution. Furthermore, financial data is analysed to illustrate the application of the proposed functions. |
| Description | Brunel Doctoral Researcher informal Math Talk |
| Form Of Engagement Activity | A talk or presentation |
| Part Of Official Scheme? | No |
| Geographic Reach | Local |
| Primary Audience | Schools |
| Results and Impact | The Brunel DR Math Talk is the type of informal talk presentations among DR fellows within the department of Mathematics and one of us presents a talk every two weeks throughout the academic year to encourage the Math research engagement. I presented an informal talk on 25th May 2021. |
| Year(s) Of Engagement Activity | 2021 |
| Description | Brunel Maths Doctoral Researchers' Symposium 2021 |
| Form Of Engagement Activity | A talk or presentation |
| Part Of Official Scheme? | No |
| Geographic Reach | Local |
| Primary Audience | Schools |
| Results and Impact | The Math Symposium 2021 took place over the course of two days on June 10-11th and the audience was made of all the doctoral researchers and lecturers from the department of Mathematics as well as involvement of a few panelists from Institute of Mathematics and its Application (IMA) with their prize-givings. I presented a poster presentation for this Symposium. |
| Year(s) Of Engagement Activity | 2021 |