Eeective Learning Requires Neuronal Remodeling of Hebbian Synapses

This paper revisits the classical neuroscience paradigm of Hebbian learning. We nd that a necessary requirement for eeective as-sociative memory learning is that the eecacies of the incoming synapses should be uncorrelated. This requirement is diicult to achieve in a robust manner by Hebbian synaptic learning, since it depends on network level information. EEective learning can yet be obtained by a neuronal process that maintains a zero sum of the incoming synaptic eecacies. This normalization drastically improves the memory capacity of associative networks, from an essentially bounded capacity to one that linearly scales with the network's size. Such neuronal normalization can be successfully carried out by activity-dependent homeostasis of the neuron's synaptic eeca-cies, which was recently observed in cortical tissue. Thus, our nd-ings strongly suggest that eeective associative learning with Heb-bian synapses alone is biologically implausible and that Hebbian synapses must be continuously remodeled by neuronally-driven regulatory processes in the brain.