The effect of heterogeneity on decorrelation mechanisms in spiking neural networks: a neuromorphic-hardware study

Correlations in neural activity can severely impair the processing of information in neural networks. In finite-size networks, correlations are however inevitable due to common presynaptic sources. Recent theoretical studies have shown that inhibitory feedback, abundant in biological neural networks, can actively suppress these shared-input correlations and thereby enable neurons to fire nearly independently. For networks of spiking neurons, the decorrelating effect of inhibitory feedback has so far been explicitly demonstrated only for homogeneous networks of neurons with linear sub-threshold dynamics. Theory, however, suggests that the effect is a general phenomenon, present in any system with inhibitory feedback, irrespective of the details of the network structure and the neuron and synapse properties. Here, we investigate the effect of network heterogeneity on correlations in sparse, random networks of inhibitory neurons with conductance-based synapses. Accelerated neuromorphic hardware is used as a user-friendly stand-alone research tool to emulate these networks. The configurability of the hardware substrate enables us to modulate the extent of network heterogeneity in a systematic manner. We selectively study the effects of shared-input and recurrent connections on correlations in synaptic inputs and spike trains. Our results confirm that shared-input correlations are actively suppressed by inhibitory feedback also in highly heterogeneous networks exhibiting broad, heavy-tailed firing-rate distributions. However, while cell and synapse heterogeneities lead to a reduction of shared-input correlations (feedforward decorrelation), feedback decorrelation is impaired as a consequence of diminished effective feedback