pynn_brainscales.brainscales2.PlasticityRule

class pynn_brainscales.brainscales2.PlasticityRule(timer: Timer, observables: Optional[Dict[str, Union[ObservablePerSynapse, ObservablePerNeuron, ObservableArray]]] = None)

Bases: object

Plasticity rule base class. Inheritance is to be used for actual implementations. Periodic timing information is provided via class Timer. The kernel implementation is required to be in the form of C++-based PPU kernel code.

__init__(timer: Timer, observables: Optional[Dict[str, Union[ObservablePerSynapse, ObservablePerNeuron, ObservableArray]]] = None)

Create a new plasticity rule with timing information.

Parameters: timer – Timer object.

Methods

`__init__`(timer[, observables])	Create a new plasticity rule with timing information.
`add_to_network_graph`(builder, …)
`generate_kernel`()	Generate plasticity rule kernel to be compiled into PPU program.
`get_data`(network_graph, outputs)	Get synaptic and neuron observables of plasticity rule.
`get_observable_array`(observable)	Get data for an array observable.

Attributes

`observables`
`timer`

class ObservableArray

Bases: pybind11_builtins.pybind11_object

class Type

Bases: pybind11_builtins.pybind11_object

class Int16: Bases: pybind11_builtins.pybind11_object

class Int8: Bases: pybind11_builtins.pybind11_object

class UInt16: Bases: pybind11_builtins.pybind11_object

class UInt8: Bases: pybind11_builtins.pybind11_object

int16 = int16

int8 = int8

uint16 = uint16

uint8 = uint8

property size

property type

class ObservablePerNeuron

Bases: pybind11_builtins.pybind11_object

class Layout

Bases: pybind11_builtins.pybind11_object

Members:

complete_row

packed_active_columns

complete_row = <Layout.complete_row: 0>

property name

packed_active_columns = <Layout.packed_active_columns: 1>

property value

class Type

Bases: pybind11_builtins.pybind11_object

class Int16: Bases: pybind11_builtins.pybind11_object

class Int8: Bases: pybind11_builtins.pybind11_object

class UInt16: Bases: pybind11_builtins.pybind11_object

class UInt8: Bases: pybind11_builtins.pybind11_object

int16 = int16

int8 = int8

uint16 = uint16

uint8 = uint8

property layout

property type

class ObservablePerSynapse

Bases: pybind11_builtins.pybind11_object

class LayoutPerRow

Bases: pybind11_builtins.pybind11_object

Members:

complete_rows

packed_active_columns

complete_rows = <LayoutPerRow.complete_rows: 0>

property name

packed_active_columns = <LayoutPerRow.packed_active_columns: 1>

property value

class Type

Bases: pybind11_builtins.pybind11_object

class Int16: Bases: pybind11_builtins.pybind11_object

class Int8: Bases: pybind11_builtins.pybind11_object

class UInt16: Bases: pybind11_builtins.pybind11_object

class UInt8: Bases: pybind11_builtins.pybind11_object

int16 = int16

int8 = int8

uint16 = uint16

uint8 = uint8

property layout_per_row

property type

add_to_network_graph(builder: _pygrenade_vx_network.NetworkBuilder, snippet_begin_time, snippet_end_time) → _pygrenade_vx_network.PlasticityRuleOnNetwork

generate_kernel() → str

Generate plasticity rule kernel to be compiled into PPU program. The interface to be adhered to is the same as in the empty implementation below. PLASTICITY_RULE_KERNEL is the generic name of the kernel function, which will be expanded to a unique implementation-defined name upon compilation to allow for multiple kernels.

Returns: PPU-code of plasticity-rule kernel as string.

get_data(network_graph: grenade.NetworkGraph, outputs: grenade_signal_flow.OutputData) → grenade.PlasticityRule.RecordingData

Get synaptic and neuron observables of plasticity rule.

Parameters

network_graph – Network graph to use for lookup of MADC output vertex descriptor.
outputs – All outputs of a single execution to extract samples from.

Returns

Recording data.

get_observable_array(observable: str) → object

Get data for an array observable.

Parameters

observable – Name of observable.

Returns

Array with recorded data. The array’s entries are values for each timer entry. Each value has a .data attribute, containing the recorded data. This data is twice the size set when initializing the observable, since it is added for both top and bottom PPUs.

Raises

RuntimeError – If observable name is not known.
TypeError – If observable is not an ObservableArray.

property observables

property timer