Temporal prediction and the mammalian auditory system

Neurons in the sensory systems are tuned to diverse but specific features of the sensory environment. Various hypotheses have been proposed to explain how this particular tuning may be the result of optimisation to statistical regularities of natural sensory stimuli under a given principle, such as efficient coding, sparse coding, and slowness. In this thesis, we investigate whether the auditory system is governed by temporal prediction. This principle states that the sensory features represented by neurons are optimised to predict immediate future input from recent past input. Experimental evidence indicates that neurons in the retina extract the most predictive information at a near-optimal efficiency level. Recent theoretical work has also demonstrated that temporal prediction can explain tuning across different sensory modalities. An artificial neural network optimised for temporal prediction of natural moving images can explain many of the receptive field features exhibited by neurons at different stages along the visual pathway. An auditory counterpart applied to spectrogram-like representations of natural sounds similarly produces receptive fields resembling those in the primary auditory cortex. In this thesis, we present work that further explores this explanatory principle for the auditory system. First, we show that an artificial neural network optimised for temporal prediction of raw waveforms of natural sounds produces tuning properties that are both qualitatively and quantitatively similar to those observed in the cochlea. Second, we show that a decision model based on auditory temporal prediction can predict the behaviour of mice in multiple auditory discrimination tasks. Finally, we present work that investigates whether a recurrent temporal prediction model can explain functional connectivity and neural responses to natural sounds in primary auditory cortex. Together, these findings provide further support for the role of temporal prediction in accounting for diverse characteristics of the auditory system