Boerlin et al.

• We've been trying to figure out a paper by Boerlin, Machens, and Deneve from 2013: Predictive Coding of Dynamical Variables in Balanced Spiking Networks.
• This can be used for robotics and other applications that follow dynamical variable.
• We're working on a simple function an integrator. This just retains how much energy has been put into the system or removed.
• One set of neurons maintain a positive value, and the other set a negative value.
• The neurons act near the edge, and communicate with fast and slow synapses.
• The fast synapses encourage the other set, and the slow synapses maintain your own set.
• You can then use a kernel to interpret the firing to tell you what the current value is.