inferno.neural

Modelling

Layer	Representation of simultaneously processed connections and neurons.
Cell	Pair of a Connection and Neuron produced used for training.
Updater	Managed accumulated updates for module parameters.
Updatable	Adds parameter updating functionality to a module.
Accumulator	Used to accumulate updates for a parameter.

Layers

Biclique	Layer structured as a complete bipartite graph.
Serial	Layer with a single connection and single neuron group.
RecurrentSerial	Layer with a single feedforward connection and neuron group, and two feedback connections with a neuron group in-between.

Components

Neuron	Base class for representing a group of neurons with a common mode of dynamics.
Synapse	Base class for representing a group of input synapses for a connection.
Connection	Base class for representing a weighted connection between two groups of neurons.

Neurons

LIF	Simulation of leaky integrate-and-fire (LIF) neuron dynamics.
ALIF	Simulation of adaptive leaky integrate-and-fire (ALIF) neuron dynamics.
GLIF1	Simulation of generalized leaky integrate-and-fire 1 (GLIF1) neuron dynamics.
GLIF2	Simulation of generalized leaky integrate-and-fire 2 (GLIF2) neuron dynamics.
QIF	Simulation of quadratic integrate-and-fire (QIF) neuron dynamics.
Izhikevich	Simulation of Izhikevich (adaptive quadratic) neuron dynamics.
EIF	Simulation of exponential integrate-and-fire (EIF) neuron dynamics.
AdEx	Simulation of adaptive exponential integrate-and-fire (AdEx) neuron dynamics.

Synapses

DeltaCurrent	Memoryless synapse which responds instantaneously to input.
DeltaPlusCurrent	Memoryless synapse which responds instantaneously to input, with passthrough current.
SingleExponentialCurrent	Instantly applied exponentially decaying current-based synapse.
DoubleExponentialCurrent	Exponentially applied exponentially decaying current-based synapse.

Connections

LinearDense	Linear all-to-all connection.
LinearDirect	Linear one-to-one connection.
LinearLateral	Linear all-to-"all but one" connection.
Conv2D	Convolutional connection along two spatial dimensions with separate input planes.

Encoders

HomogeneousPoissonEncoder	Encoder to generate spike trains sampled from a Poisson distribution.
HomogeneousPoissonApproxEncoder	Encoder to generate spike trains approximating being sampled from a Poisson distribution.
PoissonIntervalEncoder	Encoder to generate spike trains with intervals sampled from a Poisson distribution.

Hooks

Clamping	Clamps attribute of registered module on call.
Normalization	Normalizes attribute of registered module on call.

Types

SynapseConstructor

Common constructor for synapses, used by Connection objects.

Internal Components

InfernoNeuron	Base class for neurons included in the Inferno library.
InfernoSynapse	Base class for representing synapses included in the Inferno library.

Internal Mixins

mixins.BatchMixin	Mixin for modules with batch-size dependent parameters or buffers.
mixins.ShapeMixin	Mixin for modules with a concept of shape.
mixins.BatchShapeMixin	Mixin for modules with a concept of shape and with batch-size dependencies.
mixins.DelayedMixin	Mixin for modules with delay-record tensors with shared step time and duration.
neurons.mixins.AdaptiveCurrentMixin	Mixin for neurons with adaptative input currents.
neurons.mixins.AdaptiveThresholdMixin	Mixin for neurons with adaptative thresholds.
neurons.mixins.CurrentMixin	Mixin for neurons with membrane currents.
neurons.mixins.RefractoryMixin	Mixin for neurons with refractory periods.
neurons.mixins.SpikeRefractoryMixin	Mixin for neurons with refractory periods with spikes based off of them.
neurons.mixins.VoltageMixin	Mixin for neurons driven by membrane voltage.
synapses.mixins.CurrentMixin	Mixin for synapses with current primitive.
synapses.mixins.SpikeMixin	Mixin for synapses with spike primitive.
synapses.mixins.CurrentDerivedSpikeMixin	Mixin for synapses with current and spikes derived therefrom.
synapses.mixins.SpikeDerivedCurrentMixin	Mixin for synapses with spikes and currents derived therefrom.
synapses.mixins.SpikeCurrentMixin	Mixin for synapses with primitive current and spikes.
connections.mixins.WeightMixin	Mixin for connections with weights.
connections.mixins.WeightBiasMixin	Mixin for connections with weights and biases.
connections.mixins.WeightBiasDelayMixin	Mixin for connections with weights, biases, and delays.
encoders.mixins.StepTimeMixin	Mixin for encoders with a base step time.
encoders.mixins.StepMixin	Mixin for encoders with a globally meaningful number of steps.
encoders.mixins.RefractoryStepMixin	Mixin for encoders with a refractory period and a notion of global step.
encoders.mixins.GeneratorMixin	Mixin for encoders with a random number generator.