Continuous Spiking Neural Network

A continuous spiking neural network is a spiking neural network where the neurons produce continuous outputs. The refractory period after a spike is emulated using a continuous variable called a discount, which is based on the output of a neuron. A discount is subtracted from the input of a neuron making it more difficult to activate if it has recently produced a high output. The discount decays over time until it reaches zero, at which point it no longer affects the neuron.

Visualization

Example

# Create network
network=Network(
	discount_factor=0.2,
	discount_decay=0. ,
	learning_rate=0.1,
	decay_rate=0.01)

# Create nodes
network.create_nodes(
	input_count=6,
	hidden_count=15,
	output_count=5)

# Create input-to-hidden links
network.create_projection(
	sources=network.input_indices,
	targets=network.hidden_indices,
	connectivity=0.25)

# Create hidden-to-hidden links
network.create_projection(
	sources=network.hidden_indices,
	targets=network.hidden_indices,
	connectivity=0.1)

# Create hidden-to-output links
network.create_projection(
	sources=network.hidden_indices,
	targets=network.output_indices,
	connectivity=0.25)

sample=[randrange(2) for i in range(len(network.input_indices))]
output=network.update(sample)
print(output)