Focus Issue on Principles of Neural Computation for Artificial Behaving Agents

Neuromorphic processing systems implementing spiking neural networks with mixed signal analog/digital electronic circuits and/or memristive devices represent a promising technology for edge computing applications that require low power, low latency, and that cannot connect to the cloud for off-line processing, either due to lack of connectivity or for privacy concerns. However, these circuits are typically noisy and imprecise, because they are affected by device-to-device variability, and operate with extremely small currents. So achieving reliable computation and high accuracy following this approach is still an open challenge that has hampered progress on the one hand and limited widespread adoption of this technology on the other. By construction, these hardware processing systems have many constraints that are biologically plausible, such as heterogeneity and non-negativity of parameters. More and more evidence is showing that applying such constraints to artificial neural networks, including those used in artificial intelligence, promotes robustness in learning and improves their reliability. Here we delve even more into neuroscience and present network-level brain-inspired strategies that further improve reliability and robustness in these neuromorphic systems: we quantify, with chip measurements, to what extent population averaging is effective in reducing variability in neural responses, we demonstrate experimentally how the neural coding strategies of cortical models allow silicon neurons to produce reliable signal representations, and show how to robustly implement essential computational primitives, such as selective amplification, signal restoration, working memory, and relational networks, exploiting such strategies. We argue that these strategies can be instrumental for guiding the design of robust and reliable ultra-low power electronic neural processing systems implemented using noisy and imprecise computing substrates such as subthreshold neuromorphic circuits and emerging memory technologies.

In the future, artificial agents will need to make assessments of tactile stimuli in order to interact intelligently with the environment and with humans. Such assessments will depend on exquisite and robust mechanosensors, but sensors alone do not make judgments and choices. Rather, the central processing of mechanosensor inputs must be implemented with algorithms that produce 'behavioral states' in the artificial agent that resemble or mimic perceptual judgments in biology. In this study, we consider the problem of perceptual judgment as applied to vibration intensity. By a combination of computational modeling and simulation followed by psychophysical testing of vibration intensity perception in rats, we show that a simple yet highly salient judgment—is the current stimulus strong or weak?—can be explained as the comparison of ongoing sensory input against a criterion constructed as the time-weighted average of the history of recent stimuli. Simulations and experiments explore how judgments are shaped by the distribution of stimuli along the intensity dimension and, most importantly, by the time constant of integration which dictates the dynamics of criterion updating. The findings of this study imply that judgments made by the real nervous system are not absolute readouts of physical parameters but are context-dependent; algorithms of this form can be built into artificial systems.