Neural Heterogeneity Enhances Reliable Neural information Processing: Local Sensitivity and Globally Input-slaved Transient Dynamics
Neural Heterogeneity Enhances Reliable Neural information Processing: Local Sensitivity and Globally Input-slaved Transient Dynamics
Wu, S.; Huang, H.; Wang, S.; Chen, G.; Zhou, C.; Yang, D.
AbstractCortical neuronal activity exhibits variability over time and across repeated stimulation trials, yet consistently represents stimulus features. However, the dynamical mechanism underlying this reliable representation and computation remains elusive. This study uncovers a mechanism that achieves reliable neural processing, leveraging a biologically plausible network model with spatial extension and neuronal timescale diversity. Our findings demonstrate that timescale diversity disrupts intrinsic coherent spatiotemporal patterns, enhancing local sensitivity and aligning neural network activity closely with inputs. This underscores the significant role of timescale diversity in shaping consistent stimulus representation, leading to local sensitivity and globally input-slaved transient dynamics, essential for reliable neural processing. This mechanism offers a potentially general framework for understanding the role of neural heterogeneity in reliable computation. Our work informs the design of new reservoir computing models endowed with liquid wave reservoirs for neuromorphic computing.