One dimensional approximations of neuronal dynamics reveal computational strategy

The relationship between neuronal activity and computations embodied by it remains an open question. We develop a novel methodology that condenses observed neuronal activity into a quantitatively accurate, simple, and interpretable model and validate it on diverse systems and scales from single neurons in C. elegans to fMRI in humans. The model treats neuronal activity as collections of interlocking 1-dimensional trajectories. Despite their simplicity, these models accurately predict future neuronal activity and future decisions made by human participants. Moreover, the structure formed by interconnected trajectories—a scaffold—is closely related to the computational strategy of the system. We use these scaffolds to compare the computational strategy of primates and artificial systems trained on the same task to identify specific conditions under which the artificial agent learns the same strategy as the primate. The computational strategy extracted using our methodology predicts specific errors on novel stimuli. These results show that our methodology is a powerful tool for studying the relationship between computation and neuronal activity across diverse systems.Author summary: Advances in neuronal imaging techniques now allow for the recording of an appreciable fraction of neurons in biological systems. However, it is not clear how to extract scientific insight from such complex nonlinear time series data. Here we develop a methodology for the direct interpretation of the computational algorithms used by several distinct biological systems during a variety of laboratory tasks. The primary intuition of this methodology is that in order to store information in noisy dynamical systems, the information must be protected against noise. This criterion results in systems that can be well approximated by one-dimensional trajectories. These trajectories map directly to the computation performed by the system, and don’t dramatically reduce the quantitative accuracy of the models. We apply this method to several different biological systems and artificial neural networks. We find that biological systems tend to find generalistic solutions to problems, while artificial neural networks rely on specifics of the dataset.