Gradient learning in spiking neural networks by dynamic perturbation of conductances

Abstract

We present a method of estimating the gradient of an objective function with respect to the synaptic weights of a spiking neural network. The method works by measuring the fluctuations in the objective function in response to dynamic perturbation of the membrane conductances of the neurons. It is compatible with recurrent networks of conductance-based model neurons with dynamic synapses. The method can be interpreted as a biologically plausible synaptic learning rule, if the dynamic perturbations are generated by a special class of ``empiric'' synapses driven by random spike trains from an external source.

5 pages; 1 figure; submitted to PRL

Cited by in corpus (23)

• Estimating or Propagating Gradients Through Stochastic Neurons for Conditional Computation
• Estimating or Propagating Gradients Through Stochastic Neurons
• Towards an integration of deep learning and neuroscience
• Simple, Fast and Accurate Implementation of the Diffusion Approximation Algorithm for Stochastic Ion Channels with Multiple States
• STDP as presynaptic activity times rate of change of postsynaptic activity
• Learning to solve the credit assignment problem
• Exploiting Noise as a Resource for Computation and Learning in Spiking Neural Networks
• Early Inference in Energy-Based Models Approximates Back-Propagation
• Multiplexed gradient descent: Fast online training of modern datasets on hardware neural networks without backpropagation
• Structured Convolution Matrices for Energy-efficient Deep learning
• Gradient-Free Training of Recurrent Neural Networks using Random Perturbations
• Emergent Computations in Trained Artificial Neural Networks and Real Brains
• Propagating Asymptotic-Estimated Gradients for Low Bitwidth Quantized Neural Networks
• Neuron as a reward-modulated combinatorial switch and a model of learning behavior
• A nonlinear hidden layer enables actor-critic agents to learn multiple paired association navigation
• Thalamocortical contribution to solving credit assignment in neural systems
• Training recurrent neural networks with sparse, delayed rewards for flexible decision tasks
• A model of reward-modulated motor learning with parallelcortical and basal ganglia pathways
• Scaling of hardware-compatible perturbative training algorithms
• Rules and mechanisms for efficient two-stage learning in neural circuits
• Learning recurrent dynamics in spiking networks
• Statistical Mechanics of Node-perturbation Learning with Noisy Baseline
• Reinforcement learning of recurrent neural network for temporal coding