ml_genn.readouts package

Readouts are used to convert the internal state of an output neuron population into a form that can be used for classification. Unlike ANNs where this this is typically just the activation of the neurons, with SNNs it can be some function of their internal state over time or the spike times.

class ml_genn.readouts.AvgVar

Bases: Readout

Read out per-neuron average of neuron model’s output variable

add_readout_logic(model, **kwargs)

Create a copy of a neuron model with any additional state and functionality required to implement this readout added.

Parameters:: model (NeuronModel) – Base neuron model
Return type:: NeuronModel

get_readout(genn_pop, batch_size, shape)

Read out the value from the state of a compiled neuron group.

Parameters:

genn_pop – GeNN NeuronGroup object population with readout has been compiled into
batch_size (int) – Batch size of model readout is part of
shape – Shape of population

Return type:

ndarray

property reset_vars: Get list of tuples describing name, type and value to reset any state variables added by readout to

class ml_genn.readouts.AvgVarExpWeight

Bases: Readout

Read out per-neuron average of neuron model’s output variable with exponential weighting as described by [Nowotny2024].

add_readout_logic(model, **kwargs)

Create a copy of a neuron model with any additional state and functionality required to implement this readout added.

Parameters:: model (NeuronModel) – Base neuron model
Return type:: NeuronModel

get_readout(genn_pop, batch_size, shape)

Read out the value from the state of a compiled neuron group.

Parameters:

genn_pop – GeNN NeuronGroup object population with readout has been compiled into
batch_size (int) – Batch size of model readout is part of
shape – Shape of population

Return type:

ndarray

property reset_vars: Get list of tuples describing name, type and value to reset any state variables added by readout to

class ml_genn.readouts.FirstSpikeTime

Bases: Readout

Read out time of first spike emitted by population. Spike times are negated so standard metrics and loss functions can be employed.

add_readout_logic(model, **kwargs)

Create a copy of a neuron model with any additional state and functionality required to implement this readout added.

Parameters:: model (NeuronModel) – Base neuron model
Return type:: NeuronModel

get_readout(genn_pop, batch_size, shape)

Read out the value from the state of a compiled neuron group.

Parameters:

genn_pop – GeNN NeuronGroup object population with readout has been compiled into
batch_size (int) – Batch size of model readout is part of
shape – Shape of population

Return type:

ndarray

property reset_vars: Get list of tuples describing name, type and value to reset any state variables added by readout to

class ml_genn.readouts.MaxVar

Bases: Readout

Read out per-neuron maximum value of neuron model’s output variable

add_readout_logic(model, **kwargs)

Create a copy of a neuron model with any additional state and functionality required to implement this readout added.

Parameters:: model (NeuronModel) – Base neuron model
Return type:: NeuronModel

get_readout(genn_pop, batch_size, shape)

Read out the value from the state of a compiled neuron group.

Parameters:

genn_pop – GeNN NeuronGroup object population with readout has been compiled into
batch_size (int) – Batch size of model readout is part of
shape – Shape of population

Return type:

ndarray

property reset_vars: Get list of tuples describing name, type and value to reset any state variables added by readout to

class ml_genn.readouts.Readout

Bases: ABC

Base class for all readouts

abstract add_readout_logic(model, **kwargs)

Create a copy of a neuron model with any additional state and functionality required to implement this readout added.

Parameters:: model (NeuronModel) – Base neuron model
Return type:: NeuronModel

abstract get_readout(genn_pop, batch_size, shape)

Read out the value from the state of a compiled neuron group.

Parameters:

genn_pop – GeNN NeuronGroup object population with readout has been compiled into
batch_size (int) – Batch size of model readout is part of
shape – Shape of population

Return type:

ndarray

property reset_vars: List[Tuple[str, str, Number]]: Get list of tuples describing name, type and value to reset any state variables added by readout to

class ml_genn.readouts.SpikeCount

Bases: Readout

Read out number of spikes emitted by population

add_readout_logic(model, **kwargs)

Create a copy of a neuron model with any additional state and functionality required to implement this readout added.

Parameters:: model (NeuronModel) – Base neuron model
Return type:: NeuronModel

get_readout(genn_pop, batch_size, shape)

Read out the value from the state of a compiled neuron group.

Parameters:

genn_pop – GeNN NeuronGroup object population with readout has been compiled into
batch_size (int) – Batch size of model readout is part of
shape – Shape of population

Return type:

ndarray

property reset_vars: Get list of tuples describing name, type and value to reset any state variables added by readout to

class ml_genn.readouts.SumVar

Bases: Readout

Read out per-neuron sum of neuron model’s output variable

add_readout_logic(model, **kwargs)

Create a copy of a neuron model with any additional state and functionality required to implement this readout added.

Parameters:: model (NeuronModel) – Base neuron model
Return type:: NeuronModel

get_readout(genn_pop, batch_size, shape)

Read out the value from the state of a compiled neuron group.

Parameters:

genn_pop – GeNN NeuronGroup object population with readout has been compiled into
batch_size (int) – Batch size of model readout is part of
shape – Shape of population

Return type:

ndarray

property reset_vars: Get list of tuples describing name, type and value to reset any state variables added by readout to

class ml_genn.readouts.Var

Bases: Readout

Read out instantaneous value of neuron model’s output variable

add_readout_logic(model, **kwargs)

Create a copy of a neuron model with any additional state and functionality required to implement this readout added.

Parameters:: model (NeuronModel) – Base neuron model
Return type:: NeuronModel

get_readout(genn_pop, batch_size, shape)

Read out the value from the state of a compiled neuron group.

Parameters:

genn_pop – GeNN NeuronGroup object population with readout has been compiled into
batch_size (int) – Batch size of model readout is part of
shape – Shape of population

Return type:

ndarray