Fig. 1

A diagram illustrating the simulation procedure. In the beginning of a simulation, we initialize the number of neurons, SNR, the uniformity of PDs, the proportion or the uniformity of well-tuned neurons, and model parameters for the non-stationarity of PDs. Then, the PD of each neuron in an ensemble is determined based on the initial conditions. The spike train of each neuron is generated using the kinematic data generated by hand movements. The spike data and hand movement data are used to build and evaluate decoding algorithms