Low-pass filter effects in biological neurons as a feature to facilitate representation of tactile information

Here we simulate the tactile sensory information processing in a fully recurrent network of excitatory and inhibitory neurons. The neuron model has previously been shown to capture essential aspects of the Hodgkin-Huxley biological neuron model. We specifically explore the impact of noise in the spike generation of the tactile sensors, and how the time constants of the biological neurons in a network may be adapted to cope with such noise. We find that in networks with short time constants, and hence higher temporal precision, the risk is high that the network will segregate noise (i.e. as in overfitting) in the spike generation rather than the underlying haptic input signal, which is composed of features distributed across somewhat longer periods of time. Hence, low-pass filtering effects that arise because of these time constants can be beneficial for a biological neuronal network processing tactile information, to focus its available capacity on the main underlying haptic input features.