Symbolic representation of neuronal dynamics

We demonstrate a GPU-based symbolic toolkit to study a whole range of dynamical behaviors occurring in neuron models. Its algorithms include periodicity detection, hashing, and Lempel-Ziv complexity to process symbolic sequences extracted from wave-form traces using voltage and time interval partitions. This aggregated partitioning scheme is well applicable to a broad spectrum of other dynamical systems across diverse disciplines. Our approach is motivated by experimental neurophysiology where voltage wave-forms are often the only observables available. This symbolic toolkit can offset and complement other computational tools for studying neuronal dynamics such as spike counting, Lyapunov exponents and parameter continuation.