Brain-inspired machine learning for optimizing deep brain stimulation
This project lies on the interface of machine learning and neuroscience. Each of these areas can benefit from the insight from the other. For example, machine learning can be used for optimizing treatments for neurological disorder, while understanding the brain can help develop new machine learning algorithms. Deep brain stimulation (DBS) is a treatment for several neurological conditions, such as Parkinson’s disease and essential tremor. In these disorders, there are abnormal oscillations in neural activity, and some of them result in the characteristic shaking of hands. DBS desynchronizes these abnormal oscillations. Our group has developed mathematical models of the effects of DBS which suggest how to optimally control DBS to best supress the neural oscillations (Weerasinghe et al. 2019, 2021, PLoS Computational Biology). However, our theory assumes knowledge of certain parameters of the neural system, which need to be estimated. Thus, machine learning may help to find best control policies for DBS. Our group has also investigated the relationship between deep learning algorithms and model of learning in the brain (Whittington & Bogacz, 2019, Trends in Neuroscience). We recently observed that certain biologically inspired models learn more effectively than standard artificial neural networks.
The goal of the project is to develop machine learning methods to find optimal control policy for DBS. The first stage will involve numerical experiments investigating if machine learning algorithms can discover the optimal control policies for DBS in a simulated neural system. Next, it will be tested how these machine learning algorithms perform in experiments with real DBS. The work on the project will involve development and analysis of mathematical models, computer simulations, and comparison with experimental data gathered in BNDU and by other collaborators in Oxford. The applicants need to have strong mathematical skills and background (in calculus, linear algebra and probability theory), and be proficient in computer programming
The project will take place in the Medical Research Council Brain Network Dynamics Unit at the University of Oxford and students will benefit from the both the extensive generic and transdisciplinary skills training available within the Unit. This particular project will also offer specific training in mathematical modelling, computer simulations, data analysis and machine learning.
This Ph.D. (D.Phil.) studentship is funded by the Medical Research Council (MRC), a part of UKRI. The successful applicant is entitled to receive a tax-free stipend and, as a minimum, tuition fees paid at the Home level, regardless of whether they are Home or International students. Please see further details about MRC/UKRI studentships and guidance regarding Home and International eligibility. International students may be charged international level (‘Overseas’) fees by the University. Overseas fees are typically higher than Home level fees. See further details on fees at the University. Note that this MRC-funded studentship is not able to cover tuition fees above the Home level.
Interested candidates should possess, or expect to receive, a 1st class or upper 2nd class degree (or equivalent) in a related scientific discipline, e.g. computer science, engineering, mathematics, biological or physical sciences. Strong mathematical skills and background (in calculus, linear algebra and probability theory) are essential, as is proficiency in computer programming.
Prof Rafal Bogacz will be happy to discuss the project and PhD further. Please contact him by email on email@example.com
To be considered for one of these MRC-funded studentships, please submit an application for admission to the D.Phil. in the Nuffield Department of Clinical Neurosciences (course code RD_CU1), following the guidance. According to this guidance, the application should include a brief research proposal, which in this case should describe your ideas on how machine learning could be used to optimize closed-loop deep brain stimulation. On the application form, in the section headed ‘Departmental Studentship Applications’, please indicate that you are applying for a studentship and enter the reference code for a MRC BNDU studentship "2022BNDUDTA".
The closing date for applications is 12.00 midday UK time on Friday 3rd December 2021.
Interviews for short-listed applicants will be held on Friday 7th and Monday 10th January 2022.
Applications are invited from both Home students and International students to join a multidisciplinary team of scientists studying computational models of brain decision networks. An MRC-funded studentship is available from the start of academic year 2022/23, for 3.5 years, and will be primarily supervised by Professor Rafal Bogacz at the MRC BNDU.