PrincetonUniversity
bidirectional interaction?
How can I implement bi-directional projections in PsyNeuLink (e.g. the kind of projections used in the Interactive Activation and Competition model)? I want to get a recurrent system that can do pattern completion. It would be great if there are some working examples, but some high-level suggestions would be good enough! Thanks!
There are several ways. The easiest is to use an existing Mechanism. The most general is a RecurrentTransferMechanism. There is also a subclass of that, LCA (for leaky competitive accumulator) that implements the Usher & McClelland (2001) model. Finally, there is an implementation of a Leabra mechanism. That is less well tested, and has known limitations, but if you specifically wanted to include that learning algorithm (for hcmp simulations), then that might be a way to go.
jdc
On Dec 1, 2017, at 1:58 PM, Qihong Lu [email protected] wrote:
How can I implement bi-directional projection in PsyNeuLink (e.g. the kind of projections used in the Interactive Activation and Competition https://web.stanford.edu/group/pdplab/pdphandbook/handbookch3.html model)? I want to get a recurrent system that can do pattern completion. It would be great if there are some working examples, but some high-level suggestions would be good enough! Thanks!
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/PrincetonUniversity/PsyNeuLink/issues/562, or mute the thread https://github.com/notifications/unsubscribe-auth/AKR1GJBsq2SzK216geKg_c5tzNTHg1kKks5s8ExTgaJpZM4QyxFz.
