The LMU is a particular type of recurrent model that aims to optimally remember time-series information in the form of coefficients of Legendre Polynomials.
A Nengo example showing how to implement the LMU can be found here. Each LMU will only be able to remember a single dimension, so we can take our neurons, randomly project to some lower dimension, and implement an LMU for each dimension -- the LMU output can then be used to decode the output.
The LMU is a particular type of recurrent model that aims to optimally remember time-series information in the form of coefficients of Legendre Polynomials.
A Nengo example showing how to implement the LMU can be found here. Each LMU will only be able to remember a single dimension, so we can take our neurons, randomly project to some lower dimension, and implement an LMU for each dimension -- the LMU output can then be used to decode the output.