Neuron Types and Parameters

Neuron Types and Parameters:

As of now the model supports three types of neurons:

basic - Standard Leaky integrate and fire neurons.

These neurons are the ones should probably be using the most often.

Parameters:

time_constant, sigmoid_center, sigmoid_slope, refractory_voltage

• time_constant: The time constant affects how rapidly the neuron's voltage will change. This change can either be the leak, or the change resulting from another neuron synapsed onto this neuron. Note that this does not effect external stimuli.
  • Leak: they have a fixed resting voltage set to 0. The voltage of an isolated neuron will gradually return to 0 over time.
• Probabilistic firing: These neurons have a probability of firing, which is determined by their voltage as well as by the two parameters:
  • sigmoid_center: If the voltage of the neuron is equal to the sigmoid_center, the neuron has a 50% chance of firing. It can be thought of as the firing threshold.
  • sigmoid_slope: Determines how random neuron firing is. The higher the sigmoid_slope, the more random the firing. When the sigmoid_slope is 0, the neuron will always fire if and only if its voltage is above the sigmoid_center.
• refractory_voltage: When the neuron fires, its voltage will immediately return to the refractory_voltage.

Default values:

• time_constant: 1.0
• sigmoid_center: 0.0
• sigmoid_slope: 0.0
• refractory_voltage: 0.0

calcium_pyr - Calcium gated pyramidal leaky integrate and fire neurons.

These neurons are a variant of the basic neurons. However, when they fire a lot, or very often, it becomes more and more difficult for these neurons to fire. When the neuron stops firing, this difficulty gradually reduces to its natural value.

Parameters:

time_constant, refractory_voltage, sigmoid_slope, sigmoid_center, calcium_inhibition_sample_t, calcium_inhibition_slope

• time_constant: The time constant affects how rapidly the neuron's voltage will change. This change can either be the leak, or the change resulting from another neuron synapsed onto this neuron. Note that this does not effect external stimuli.
  • Leak: they have a fixed resting voltage set to 0. The voltage of an isolated neuron will gradually return to 0 over time.
• Probabilistic firing: These neurons have a probability of firing, which is determined by their voltage as well as by the two parameters:
  • sigmoid_center: If the voltage of the neuron is equal to the sigmoid_center, the neuron has a 50% chance of firing. It can be thought of as the firing threshold. If calcium inhibition is activated, the sigmoid center is temporarily raised.
  • sigmoid_slope: Determines how random neuron firing is. The higher the sigmoid_slope, the more random the firing. When the sigmoid_slope is 0, the neuron will always fire if and only if its voltage is above the sigmoid_center.
• refractory_voltage: When the neuron fires, its voltage will immediately return to the refractory_voltage.
• Calcium inhibition: If these neurons fire frequently, it becomes more difficult for them to fire again. This depends on:
  • calcium_inhibition_sample_t: This integer parameter determines how many previous time steps should be considered when determining the calcium inhibition. ie: If equal to 100, the neuron will use all spikes in the last 100 time steps to calculate calcium inhibition.
  • calcium_inhibition_slope: The strength of the inhibition depends inversely on this value. ie: If the calcium_inhibition_slope is increased, the amount of inhibition decreases. This value must not be 0.

Default values:

• time_constant: 1.0
• sigmoid_center: 0.0
• sigmoid_slope: 0.0
• refractory_voltage: 0.0
• calcium_inhibition_sample_t: 1000
• calcium_inhibition_slope: 25.0

calcium_pyr_inhib - Calcium gated pyramidal leaky integrate and fire neurons.

An alternate implementation on calcium gated neurons, in which the neurons self-inhibit every time they fire.

These neurons are a variant of the basic neurons. However, when they fire a lot, or very often, it becomes more and more difficult for these neurons to fire. When the neuron stops firing, this difficulty gradually reduces to its natural value.

Parameters:

time_constant, refractory_voltage, sigmoid_slope, sigmoid_center, calcium_inhibition_sample_t, calcium_inhibition_slope

• time_constant: The time constant affects how rapidly the neuron's voltage will change. This change can either be the leak, or the change resulting from another neuron synapsed onto this neuron. Note that this does not effect external stimuli.
  • Leak: they have a fixed resting voltage set to 0. The voltage of an isolated neuron will gradually return to 0 over time.
• Probabilistic firing: These neurons have a probability of firing, which is determined by their voltage as well as by the two parameters:
  • sigmoid_center: If the voltage of the neuron is equal to the sigmoid_center, the neuron has a 50% chance of firing. It can be thought of as the firing threshold. If calcium inhibition is activated, the sigmoid center is temporarily raised.
  • sigmoid_slope: Determines how random neuron firing is. The higher the sigmoid_slope, the more random the firing. When the sigmoid_slope is 0, the neuron will always fire if and only if its voltage is above the sigmoid_center.
• refractory_voltage: When the neuron fires, its voltage will immediately return to the refractory_voltage.
• Calcium inhibition: If these neurons fire frequently, it becomes more difficult for them to fire again. Here, you can think of this as self inhibition, as if these neurons had inhibitory synapses from themselves onto themselves. This depends on:
  • ca_inhib_relevant_spike_no: This is an efficiency parameter, the higher this is the more spikes are counted in the implementation.
  • ca_inhib_tau1, ca_inhib_tau2: These parameters are equivalent to tau1 and tau2 in basic synapse, and represent how long the inhibition lasts and when it is strongest.
  • ca_inhib_weight: This parameter is equivalent to initial_weight in basic synapse, and represents how strong the inhibition is. This value should be positive, and this weight never changes.
  • ca_inhib_E: This parameter is equivalent to the nernst potential E in basic synapse, and represents the voltage that the calcium inhibition will try to pull this neuron's voltage towards. This value should be negative.

Default values:

• time_constant: 1.0
• sigmoid_center: 0.0
• sigmoid_slope: 0.0
• refractory_voltage: 0.0
• ca_inhib_relevant_spike_no: 20
• ca_inhib_tau1: 10.
• ca_inhib_tau2: 20.
• ca_inhib_weight: 0.1
• ca_inhib_E: -10

nsn - Non spiking neurons.

These neurons do not spike but instead output a linear function of their voltage at all times.
Note that these neurons must be used with non spiking synapses. (all synapses from this neuron must be non spiking)

Parameters:

time_constant

• time_constant: The time constant affects how rapidly the neuron's voltage will change. This change can either be the leak, or the change resulting from another neuron synapsed onto this neuron. Note that this does not effect external stimuli.
  • Leak: they have a fixed resting voltage set to 0. The voltage of an isolated neuron will gradually return to 0 over time.

Default values:

• time_constant: 1.0