| Paper: | PS-1A.16 | ||
| Session: | Poster Session 1A | ||
| Location: | H Lichthof | ||
| Session Time: | Saturday, September 14, 16:30 - 19:30 | ||
| Presentation Time: | Saturday, September 14, 16:30 - 19:30 | ||
| Presentation: | Poster | ||
| Publication: | 2019 Conference on Cognitive Computational Neuroscience, 13-16 September 2019, Berlin, Germany | ||
| Paper Title: | Elucidating Cognitive Processes Using LSTMs | ||
| Manuscript: | Click here to view manuscript | ||
| License: |  This work is licensed under a Creative Commons Attribution 3.0 Unported License. |
||
| DOI: | https://doi.org/10.32470/CCN.2019.1201-0 | ||
| Authors: | Pedro F. da Costa, Birkbeck, University of London, United Kingdom; Sebastian Popescu, Imperial College London, United Kingdom; Robert Leech, King's College London, United Kingdom; Romy Lorenz, University of Cambridge, United Kingdom | ||
| Abstract: | Despite several decades of functional neuroimaging research the relationship between brain networks and cognition remains elusive. This is because the taxonomy of cognitive processes was developed largely blind to the functional organization of the human brain. In this work, we leverage recent advances in artificial neural networks to gain insights into shared cognitive processes among six different cognitive tasks. We trained a single recurrent neural networks (RNN) to perform cognitive tasks. In this manner, we were able to evaluate shared representations between multiple cognitive tasks without relying on predefined cognitive processes. Next, we tested if the learned representations provide a good explanation for human brain activation patterns associated with these tasks. While we found little similarity between the RNN’s learned representation and real brain data, our approach offers a roadmap to gain more mechanistic insights into how cognitive processes map to brain networks with potential important implications for studying cognitive dysfunction in disease. | ||