Complete the following questionnaire for - IC2023_05 PhD position in Predoctoral Fellow in machine learning and computational neuroscience
(231 KB)
Transformers as nonequilibrium neural network models of hippocampal spatial maps. Recently, striking similarities have been drawn between models of spatial memory in the hippocampus and transformer models in deep learning. Transformers are highly successful models using self-attention mechanisms for weighting contextual information in sequential processing, leading to impressive performances in applications like large language models (BERT, GPT-3). Recent findings also show that transformers can be described as asymmetric versions of modern Hopfield neural networks, which promises to facilitate their integration with neuroscientific models. This project will exploit the connection between transformers, hippocampal models and asymmetric (nonequilibrium) Hopfield networks
Deadline: June 9th, 2023 14:00 CET
(
) Compulsory field.
The reference letters shall be submitted directly by the referees. Each referee you indicate in this form shall automatically receive an e-mail with instructions to submit the recommendation letter to BCAM. So, you DON'T have to submit yourself. Those recommendation letters sent directly by you shall automatically discarded.
If you wish to obtain further information about your rights, we recommend that you closely read the Privacy Policy
NOTICE: Once you have applied to the position, you may not re-apply in this position. Please make sure that everything is correct.
