Learning Precisely Timed Spikes

Abstract

To signal the onset of salient sensory features or execute well-timed motor sequences, neuronal circuits must transform streams of incoming spike trains into precisely timed firing. To address the efficiency and fidelity with which neurons can perform such computations, we developed a theory to characterize the capacity of feedforward networks to generate desired spike sequences. We find the maximum number of desired output spikes a neuron can implement to be 0.1-0.3 per synapse. We further present a biologically plausible learning rule that allows feedforward and recurrent networks to learn multiple mappings between inputs and desired spike sequences. We apply this framework to reconstruct synaptic weights from spiking activity and study the precision with which the temporal structure of ongoing behavior can be inferred from the spiking of premotor neurons. This work provides a powerful approach for characterizing the computational and learning capacities of single neurons and neuronal circuits.

SuperSpike: Supervised Learning in Multilayer Spiking Neural Networks

“…Similarly, Gardner and Grüning (2016) and Albers, Westkott, and Pawelzik (2016) have studied the convergence properties of rules that reduce the van Rossum distance by gradient descent. Moreover, Memmesheimer, Rubin, Ölveczky, & Sompolinsky (2014) proposed a learning algorithm that achieves high capacity in learning long, precisely timed spike trains in single units and recurrent networks. The problem of sequence learning in recurrent neural networks has also been studied as a variational learning problem (Brea, Senn, & Pfister, 2013; Jimenez Rezende & Gerstner, 2014) and by combining adaptive control theory with heterogeneous neurons (Gilra & Gerstner, 2017).…”

Questioning the role of sparse coding in the brain

“…The main reason for the interest in this coding scheme is that sparse coding can have beneficial features for both memory capacity [1,2] and speed of learning [3], as illustrated in various simulations of brain circuitry function [4][5][6][7][8][9]. From a simulation point of view, it provides the additional advantage of reduced computational load, and is therefore popular in technical applications that for example involve artificial neural networks [10,11].…”

Learning network structures from firing patterns

“…synaptic delays, which are very difficult to obtain. Additionally, Memmesheimer et al [22] proposed an inference algorithm based on the Perceptron learning rule, similar to Baldassi et al [23], for which they proved that under accurate spike times it identifies a simple n-to-1 feed forward network. They also proposed a heuristic extension that works with finite precision in recorded spike times.…”