An Internal Model Architecture for Novelty Detection: Implications for Cerebellar and Collicular Roles in Sensory Processing

Abstract

The cerebellum is thought to implement internal models for sensory prediction, but details of the underlying circuitry are currently obscure. We therefore investigated a specific example of internal-model based sensory prediction, namely detection of whisker contacts during whisking. Inputs from the vibrissae in rats can be affected by signals generated by whisker movement, a phenomenon also observable in whisking robots. Robot novelty-detection can be improved by adaptive noise-cancellation, in which an adaptive filter learns a forward model of the whisker plant that allows the sensory effects of whisking to be predicted and thus subtracted from the noisy sensory input. However, the forward model only uses information from an efference copy of the whisking commands. Here we show that the addition of sensory information from the whiskers allows the adaptive filter to learn a more complex internal model that performs more robustly than the forward model, particularly when the whisking-induced interference has a periodic structure. We then propose a neural equivalent of the circuitry required for adaptive novelty-detection in the robot, in which the role of the adaptive filter is carried out by the cerebellum, with the comparison of its output (an estimate of the self-induced interference) and the original vibrissal signal occurring in the superior colliculus, a structure noted for its central role in novelty detection. This proposal makes a specific prediction concerning the whisker-related functions of a region in cerebellar cortical zone A2 that in rats receives climbing fibre input from the superior colliculus (via the inferior olive). This region has not been observed in non-whisking animals such as cats and primates, and its functional role in vibrissal processing has hitherto remained mysterious. Further investigation of this system may throw light on how cerebellar-based internal models could be used in broader sensory, motor and cognitive contexts.

The Robot Vibrissal System: Understanding Mammalian Sensorimotor Co-ordination Through Biomimetics

“…ÔSignal conditioningÕ indicates the importance of early processing of sensory signals to, for instance, distinguish components of the signal that may be due to the organismÕs (or robotÕs) own movement rather than to contact with the external world. Some neurobiological evidence, and our own modelling work, suggest an important role for the cerebellum in this regard (Anderson, Pearson et al, 2010;Anderson, Porrill et al, 2012).…”

“…In a further extension of our sensory noise cancellation model (Anderson, Porrill et al, 2012) we have shown that the addition of sensory information from the whiskers allows the adaptive filter to learn a more complex internal model that performs more robustly than a forward model based on efference copy signals alone, particularly when the whisking-induced interference has a periodic structure. More generally, our analysis of the whisking noise cancellation scheme reveals that the functional role of the cerebellum may be to learn a forward model of the whisker/follicle dynamics.…”

“…This links to separate speculation over the functional role of the cerebellum in motor control and sensory processing, where it has been suggested that the cerebellum can learn a variety of forward and inverse models in control and state estimation tasks, see for instance (Dean and Porrill, 2014;Wolpert, Miall et al, 1998). Our development of the whisking noise cancellation scheme from a theoretical basis has led to a number of experimental predictions relating to the functionality of different components of the cerebellar micro-circuit: (i) that the mossy fibres transmit a copy of motor command, (ii) that the Purkinje cell output is an estimate of the self-induced noise signal, (iii) that the climbing fibre teaching signal is an estimate of the ÔcleanÕ whisker sensory signal, and (iv) that the superior colliculus is the target of the cerebellar output and acts to compare predicted and actual sensory signals (Anderson, Porrill et al, 2012).…”

“…An important role might lie in the calibration of sensory maps used to generate fast orienting movements. We have hypothesised that the known extensive cerebellar-collicular connectivity (see (Anderson, Porrill et al, 2012;Teune, van der Burg et al, 2000)), together with the adaptive filter cerebellar architecture described above, could play a role in calibrating predictive topographic maps in the colliculus. We are currently investigating how this model can be employed in a predictive architecture in which features appear in the salience map at their predicted rather than their current positions.…”

Functional Role of the Cerebellum in Gamma-Band Synchronization of the Sensory and Motor Cortices

“…Parts of the cerebellar cortex process the whisker sensory information (Bosman et al, 2010) but the functional importance of this activity is still unknown. Interestingly, while the cerebellar inactivation seemed to minimally perturb whiskers' movements, a longer fraction of the total whisking time was spent exploring objects, suggesting a failure to adapt to familiar objects (Anderson et al, 2012).…”