Open Code: A Flexible Model of Working Memory

Code for simulating a computational model of flexible working memory. The model relies on random connections between a structured neural network and a random neural network to maintain representations over a delay. The randomness of the model endows it with the flexibility to remember arbitrary inputs. Although, this comes at the cost of a limited capacity. Details of the model are availabe in the paper "A Flexible Model of Working Memory" by Bouchacourt and Buschman, 2019. 

Code is written in Python. Model code, as well as example scripts for recreating figures, are available on our lab's GitHub repository.