Causal Neural Mechanisms for Decision Making: Putting Rules into Context
In nature, different rules apply under different contexts. For example, how we act in a classroom is very different from how we might be when we are with our friends. Likewise, a self-driving car would need to know that under some scenarios cautiously turning on a red light to allow an ambulance to pass may be required. However, before we can understand how we can program machines, we first need to understand how natural systems, like the brain, learn rules under different contexts. There are at least two possibilities. One is computationally inefficient but easy to implement: where every rule is learned under the relevant context. The other is where a context highlights the relevant rules that apply to it. This is a much more elegant and efficient way to achieve rule learning, and machine learning could benefit from understanding how the brain achieves this.
A motivated student has the opportunity to work with a synergistic partnership with a Newcastle lab and a lab in Durham. The Newcastle lab has built a computational model relevant for this work, and works on rule learning in primates and people. The Durham lab works with rodents on learning under different contexts. The studentship work will include designing a task based on the computational model and one that is likely to be learnable by primates and rodents. Then in Durham the task would be tested while rodents work on it with and without impairment to their hippocampal or frontal system. The Newcastle work will involve primates working on the task while an innovative approach for modulating the brain using sonic waves is used to test the role of the frontal and hippocampal systems in rule or context learning.
The potential outcome is information in two species that are crucial for translating neuroscientific information to humans and could lead to better machine learning models. The student will also receive unique skills training that are in high demand in the UK and by funders like the BBSRC. The will have an opportunity to sculpt their ideal PhD work around the topic and will be supported by a company partner (Brain Box Inc.) and a prominent figure in neuroscience who will periodically remotely join the student supervision and support meetings. The outcome could not only advance scientific knowledge on neurobiological mechanisms for crucial aspects of human intelligence but also how such mechanisms might be implemented in artificial systems.
HOW TO APPLY
Applications should be made by emailing [Email Address Removed] with a CV (including contact details of at least two academic (or other relevant) referees), and a covering letter – clearly stating your first choice project, and optionally 2nd and 3rd ranked projects, as well as including whatever additional information you feel is pertinent to your application; you may wish to indicate, for example, why you are particularly interested in the selected project(s) and at the selected University. Applications not meeting these criteria will be rejected.
In addition to the CV and covering letter, please email a completed copy of the Additional Details Form (Word document) to [Email Address Removed]. A blank copy of this form can be found at: https://www.nld-dtp.org.uk/how-apply.
Informal enquiries may be made to [Email Address Removed]
This is a 4 year BBSRC CASE studentship under the Newcastle-Liverpool-Durham DTP. The successful applicant will receive research costs, tuition fees and stipend (£15,009 for 2019-20). The PhD will start in October 2020. Applicants should have, or be expecting to receive, a 2.1 Hons degree (or equivalent) in a relevant subject. EU candidates must have been resident in the UK for 3 years in order to receive full support. Please note, there are 2 stages to the application process.
Neural mechanisms for complex combinatorial binding: Computationally and neurobiologically informed hypotheses (in press) Philosophical Transactions of the Royal Society, Biological Sciences
Changes in presynaptic calcium signalling accompany age-related deficits in hippocampal LTP and cognitive impairment. (2019) Aging Cell 18(5): e13008
The NMDA receptor antagonist MK-801 fails to impair long-term recognition memory in mice when the state-dependency of memory is controlled. (2019) Neurobiology of Learning and Memory 161: 57-62
An Open Resource for Non-human Primate Imaging. (2018) Neuron 100, 61-74.
Artificial grammar learning in vascular and progressive non-fluent aphasias. (2017)
Neuropsychologia, 104, 201-213
Sequence learning comparably modulates neuronal nested oscillations in human and monkey auditory cortex (2017). PLoS Biology e2000219.
Different forms of hierarchical effective connectivity in primate fronto-temporal pathways. (2015) Nature Communications 6, DOI: 1038/ncomms7000. Open Access
Episodic-Like Memory for What-Where-Which Occasion is Selectively Impaired in the 3xTgAD Mouse Model of Alzheimer’s Disease. (2013) Journal of Alzheimer's Disease 33(3): 681-698
Orthogonal representation of sound dimensions in the primate midbrain. (2011) Nature Neuroscience, 14(4): 423-5.