Synaptic Interference Channel

Synaptic channels automatically adapt their weights to compensate for the variations resulted from the input and output characteristics, i.e., spike frequency, time correlation among inputs, time difference between presynaptic and postsynaptic action potentials. Modification of the synaptic conductances, i.e., channel weights, is the main mechanism that enables learning in neurons. In this paper, we approach this learning mechanism from a different perspective. First, we analyze the single-input single-output (SISO) and multi-input single-output (MISO) synaptic interference channels, and achievable communication rates. Furthermore, we provide the natural adaptive weight update algorithm for neurons based on experimental findings. Our results demonstrate that neurons are capable of mitigating the interference, and achieve rates close to the capacity.