A 5-year ERC-funded project entitled “Generalisation in Mind and Machine” explores how well humans and artificial neural networks generalize across a range of tasks. Over the course of the grant there will be 4 postdoctoral researchers involved in modelling, 1 postdoctoral researcher involved in empirical work, and two PhD students. I am currently looking to hire two persons (one quite senior) with experience working with “deep” neural networks, and one person with extensive experience carrying out cognitive psychology experiments.
In contrast with the large community of modelers in machine learning focused on improving the performance of deep networks for applied reasons, the goal of this project is to explore what neural networks tell us about how the brain works. More specifically, the team will be working on questions of generalization in the domains of object identification, word identification, short-term memory, and games, amongst other areas. One key issue that we will explore is the extent to which “symbolic” computations are required to support some forms of generalization. This research is related to previous and ongoing research in my lab that explores why and when artificial neural networks learn localist "grandmother cell" representations with the goal to better understand why some neurons respond to high-level information in a highly selective manner.
The project will start in September 2017 (or shortly thereafter). For further information about the project and information about applying, please go the following links.
Research Associate focused on behavioural research: https://goo.gl/8cl6Vm
Senior Research Associate focused on modelling: https://goo.gl/u2rjk4
Research Fellow focused on modelling (most senior post): https://goo.gl/XjRKzS
Closing Date for applications: June 22, 2017. If you have any questions, please get in touch with me at: j.bowers@bristol.ac.uk
No comments:
Post a Comment