Computation-Through-Dynamics Benchmark

Computation-Through-Dynamics Benchmark Figure


This git repo contains code that will allow users to perform four basic steps:

  1. Train task-trained models on a range of tasks with different complexities
  2. Simulate synthetic neural spiking from those task-trained networks
  3. Train data-trained models on the synthetic spiking activity
  4. Compare the dynamics of the task-trained and data-trained networks with a variety of quantifications of dynamical accuracy