devices.dynapse.DynapSimCore

Bases: DynapSimCurrents, DynapSimLayout, DynapSimWeightBits

DynapSimCore stores the simulation currents and manages the conversion from configuration objects. It also provides easy update mechanisms using coarse&fine values, high-level parameter representations and etc.

Device -> Simulation current (pseudo-code)

simcore = DynapSimCore.from_Dynapse2Core(config.chips[0].cores[0])
Itau_ampa = simcore.Itau_ampa

Attributes overview

`C_ahp`	AHP synapse capacitance in Farads
`C_ampa`	AMPA synapse capacitance in Farads
`C_gaba`	GABA synapse capacitance in Farads
`C_mem`	neuron membrane capacitance in Farads
`C_nmda`	NMDA synapse capacitance in Farads
`C_pulse`	pulse-width creation sub-circuit capacitance in Farads
`C_pulse_ahp`	spike frequency adaptation circuit pulse-width creation sub-circuit capacitance in Farads
`C_ref`	refractory period sub-circuit capacitance in Farads
`C_shunt`	SHUNT synapse capacitance in Farads
`Idc`	Constant DC current injected to membrane in Amperes
`If_nmda`	NMDA gate soft cut-off current setting the NMDA gating voltage in Amperes
`Igain_ahp`	gain bias current of the spike frequency adaptation block in Amperes
`Igain_ampa`	gain bias current of excitatory AMPA synapse in Amperes
`Igain_gaba`	gain bias current of inhibitory GABA synapse in Amperes
`Igain_mem`	gain bias current for neuron membrane in Amperes
`Igain_nmda`	gain bias current of excitatory NMDA synapse in Amperes
`Igain_shunt`	gain bias current of the inhibitory SHUNT synapse in Amperes
`Io`	Dark current in Amperes that flows through the transistors even at the idle state
`Ipulse`	bias current setting the pulse width for neuron membrane `t_pulse` in Amperes
`Ipulse_ahp`	bias current setting the pulse width for spike frequency adaptation block `t_pulse_ahp` in Amperes
`Iref`	bias current setting the refractory period `t_ref` in Amperes
`Ispkthr`	spiking threshold current, neuron spikes if \(I_{mem} > I_{spkthr}\) in Amperes
`Itau_ahp`	Spike frequency adaptation leakage current setting the time constant `tau_ahp` in Amperes
`Itau_ampa`	AMPA synapse leakage current setting the time constant `tau_ampa` in Amperes
`Itau_gaba`	GABA synapse leakage current setting the time constant `tau_gaba` in Amperes
`Itau_mem`	Neuron membrane leakage current setting the time constant `tau_mem` in Amperes
`Itau_nmda`	NMDA synapse leakage current setting the time constant `tau_nmda` in Amperes
`Itau_shunt`	SHUNT synapse leakage current setting the time constant `tau_shunt` in Amperes
`Iw`	Weight bits stacked together
`Iw_0`	weight bit 0 current of the neurons of the core in Amperes
`Iw_1`	weight bit 1 current of the neurons of the core in Amperes
`Iw_2`	weight bit 2 current of the neurons of the core in Amperes
`Iw_3`	weight bit 3 current of the neurons of the core in Amperes
`Iw_ahp`	spike frequency adaptation weight current of the neurons of the core in Amperes
`Ut`	Thermal voltage in Volts
`Vth`	The cut-off Vgs potential of the transistors in Volts (not type specific)
`currents`	currents returns a subset of object which belongs to DynapSimCurrents
`gain`	Igain_ahp, Igain_ampa, Igain_gaba, Igain_nmda, Igain_shunt, Igain_mem
`kappa_n`	Subthreshold slope factor (n-type transistor)
`kappa_p`	Subthreshold slope factor (p-type transistor)
`layout`	layout returns a subset of object which belongs to DynapSimLayout
`time`	time creates the high level time constants set by currents Ipulse_ahp, Ipulse, Iref, Itau_ahp, Itau_ampa, Itau_gaba, Itau_nmda, Itau_shunt, Itau_mem
`weight_bits`	weight_bits returns a subset of object which belongs to DynapSimWeightBits

Methods overview

`__init__`([Iw_0, Iw_1, Iw_2, Iw_3, C_ahp, ...])
`compare`(core1, core2)	compare compares two DynapSimCore objects detects the different values set
`export_Dynapse2Parameters`()	export_Dynapse2Parameters converts all current values to their coarse-fine value representations for device configuration
`from_Dynapse2Core`(core)	from_Dynapse2Core is a class factory method which uses samna configuration objects to extract the simulation currents
`from_specification`([Idc, If_nmda, ...])	from_specification is a class factory method helping DynapSimCore object construction using higher level representaitons of the currents like gain ratio or time constant whenever applicable.
`get_full`(size)	get_full creates a dictionary with respect to the object, with arrays of current values
`update`(attr, value)	update_current updates an attribute and returns a new object, does not change the original object.
`update_gain_ratio`(attr, value)	update_gain_ratio updates currents setting gain ratio (Igain/Itau) attributes
`update_time_constant`(attr, value)	update_time_constant updates currents setting time constant attributes

C_ahp: float | ndarray | Tensor | array = 4e-11: AHP synapse capacitance in Farads

C_ampa: float | ndarray | Tensor | array = 2.45e-11: AMPA synapse capacitance in Farads

C_gaba: float | ndarray | Tensor | array = 2.5e-11: GABA synapse capacitance in Farads

C_mem: float | ndarray | Tensor | array = 3e-12: neuron membrane capacitance in Farads

C_nmda: float | ndarray | Tensor | array = 2.5e-11: NMDA synapse capacitance in Farads

C_pulse: float | ndarray | Tensor | array = 5e-13: pulse-width creation sub-circuit capacitance in Farads

C_pulse_ahp: float | ndarray | Tensor | array = 5e-13: spike frequency adaptation circuit pulse-width creation sub-circuit capacitance in Farads

C_ref: float | ndarray | Tensor | array = 1.5e-12: refractory period sub-circuit capacitance in Farads

C_shunt: float | ndarray | Tensor | array = 2.45e-11: SHUNT synapse capacitance in Farads

Idc: float | ndarray | Tensor | array = 5e-13: Constant DC current injected to membrane in Amperes

If_nmda: float | ndarray | Tensor | array = 5e-13: NMDA gate soft cut-off current setting the NMDA gating voltage in Amperes

Igain_ahp: float | ndarray | Tensor | array = 2.8368794326241126e-11: gain bias current of the spike frequency adaptation block in Amperes

Igain_ampa: float | ndarray | Tensor | array = 8.687943262411346e-09: gain bias current of excitatory AMPA synapse in Amperes

Igain_gaba: float | ndarray | Tensor | array = 8.865248226950353e-09: gain bias current of inhibitory GABA synapse in Amperes

Igain_mem: float | ndarray | Tensor | array = 2.1276595744680848e-11: gain bias current for neuron membrane in Amperes

Igain_nmda: float | ndarray | Tensor | array = 8.865248226950353e-09: gain bias current of excitatory NMDA synapse in Amperes

Igain_shunt: float | ndarray | Tensor | array = 8.687943262411346e-09: gain bias current of the inhibitory SHUNT synapse in Amperes

Io: float | ndarray | Tensor | array = 5e-13: Dark current in Amperes that flows through the transistors even at the idle state

Ipulse: float | ndarray | Tensor | array = 3.4999999999999996e-08: bias current setting the pulse width for neuron membrane t_pulse in Amperes

Ipulse_ahp: float | ndarray | Tensor | array = 3.5e-07: bias current setting the pulse width for spike frequency adaptation block t_pulse_ahp in Amperes

Iref: float | ndarray | Tensor | array = 1.0499999999999999e-09: bias current setting the refractory period t_ref in Amperes

Ispkthr: float | ndarray | Tensor | array = 1e-07: spiking threshold current, neuron spikes if \(I_{mem} > I_{spkthr}\) in Amperes

Itau_ahp: float | ndarray | Tensor | array = 2.8368794326241126e-11: Spike frequency adaptation leakage current setting the time constant tau_ahp in Amperes

Itau_ampa: float | ndarray | Tensor | array = 8.687943262411346e-11: AMPA synapse leakage current setting the time constant tau_ampa in Amperes

Itau_gaba: float | ndarray | Tensor | array = 8.865248226950353e-11: GABA synapse leakage current setting the time constant tau_gaba in Amperes

Itau_mem: float | ndarray | Tensor | array = 5.319148936170212e-12: Neuron membrane leakage current setting the time constant tau_mem in Amperes

Itau_nmda: float | ndarray | Tensor | array = 8.865248226950353e-11: NMDA synapse leakage current setting the time constant tau_nmda in Amperes

Itau_shunt: float | ndarray | Tensor | array = 8.687943262411346e-11: SHUNT synapse leakage current setting the time constant tau_shunt in Amperes

property Iw: ndarray: Weight bits stacked together

Iw_0: float | ndarray | Tensor | array = 1e-09: weight bit 0 current of the neurons of the core in Amperes

Iw_1: float | ndarray | Tensor | array = 2e-09: weight bit 1 current of the neurons of the core in Amperes

Iw_2: float | ndarray | Tensor | array = 4e-09: weight bit 2 current of the neurons of the core in Amperes

Iw_3: float | ndarray | Tensor | array = 8e-09: weight bit 3 current of the neurons of the core in Amperes

Iw_ahp: float | ndarray | Tensor | array = 5e-13: spike frequency adaptation weight current of the neurons of the core in Amperes

Ut: float | ndarray | Tensor | array = 0.025: Thermal voltage in Volts

Vth: float | ndarray | Tensor | array = 0.7: The cut-off Vgs potential of the transistors in Volts (not type specific)

__init__(Iw_0: float | ndarray | Tensor | array = 1e-09, Iw_1: float | ndarray | Tensor | array = 2e-09, Iw_2: float | ndarray | Tensor | array = 4e-09, Iw_3: float | ndarray | Tensor | array = 8e-09, C_ahp: float | ndarray | Tensor | array = 4e-11, C_ampa: float | ndarray | Tensor | array = 2.45e-11, C_gaba: float | ndarray | Tensor | array = 2.5e-11, C_nmda: float | ndarray | Tensor | array = 2.5e-11, C_pulse_ahp: float | ndarray | Tensor | array = 5e-13, C_pulse: float | ndarray | Tensor | array = 5e-13, C_ref: float | ndarray | Tensor | array = 1.5e-12, C_shunt: float | ndarray | Tensor | array = 2.45e-11, C_mem: float | ndarray | Tensor | array = 3e-12, Io: float | ndarray | Tensor | array = 5e-13, kappa_n: float | ndarray | Tensor | array = 0.75, kappa_p: float | ndarray | Tensor | array = 0.66, Ut: float | ndarray | Tensor | array = 0.025, Vth: float | ndarray | Tensor | array = 0.7, Idc: float | ndarray | Tensor | array = 5e-13, If_nmda: float | ndarray | Tensor | array = 5e-13, Igain_ahp: float | ndarray | Tensor | array = 2.8368794326241126e-11, Igain_ampa: float | ndarray | Tensor | array = 8.687943262411346e-09, Igain_gaba: float | ndarray | Tensor | array = 8.865248226950353e-09, Igain_nmda: float | ndarray | Tensor | array = 8.865248226950353e-09, Igain_shunt: float | ndarray | Tensor | array = 8.687943262411346e-09, Igain_mem: float | ndarray | Tensor | array = 2.1276595744680848e-11, Ipulse_ahp: float | ndarray | Tensor | array = 3.5e-07, Ipulse: float | ndarray | Tensor | array = 3.4999999999999996e-08, Iref: float | ndarray | Tensor | array = 1.0499999999999999e-09, Ispkthr: float | ndarray | Tensor | array = 1e-07, Itau_ahp: float | ndarray | Tensor | array = 2.8368794326241126e-11, Itau_ampa: float | ndarray | Tensor | array = 8.687943262411346e-11, Itau_gaba: float | ndarray | Tensor | array = 8.865248226950353e-11, Itau_nmda: float | ndarray | Tensor | array = 8.865248226950353e-11, Itau_shunt: float | ndarray | Tensor | array = 8.687943262411346e-11, Itau_mem: float | ndarray | Tensor | array = 5.319148936170212e-12, Iw_ahp: float | ndarray | Tensor | array = 5e-13) → None

__update_high_level(obj: DynapSimCoreHigh, attr_getter: Callable[[str], Any], attr: str, value: Any) → DynapSimCore

__update_high_level updates high level representations of the current values like time constants and gain ratios. The current values are updated accordingly without changing the original object.

Parameters:

obj (DynapSimCoreHigh) – the high level object that stores the projections of the current values
attr_getter (Callable[[str], Any]) – a function to get the high level attribute from the high level object
attr (str) – any attribute that belongs to any DynapSimCoreHigh object
value (Any) – the new value to set

Returns:

updated DynapSimCore object

Return type:

DynapSimCore

static compare(core1: DynapSimCore, core2: DynapSimCore) → Dict[str, Tuple[Any]][source]

compare compares two DynapSimCore objects detects the different values set

Parameters:

core1 (DynapSimCore) – the first core object
core2 (DynapSimCore) – the second core object to compare against the first one

Returns:

a dictionary of changed values

Return type:

Dict[str, Tuple[Any]]

property currents: DynapSimCurrents: currents returns a subset of object which belongs to DynapSimCurrents

export_Dynapse2Parameters() → Dict[str, Tuple[uint8, uint8]][source]

export_Dynapse2Parameters converts all current values to their coarse-fine value representations for device configuration

Returns:: a dictionary of mapping between parameter names and respective coarse-fine values
Return type:: Dict[str, Tuple[np.uint8, np.uint8]]

classmethod from_Dynapse2Core(core: Dynapse2Core) → DynapSimCore[source]

from_Dynapse2Core is a class factory method which uses samna configuration objects to extract the simulation currents

Returns:: a dynapse core simulation object whose parameters are imported from a samna configuration object
Return type:: DynapSimCore

from_specification is a class factory method helping DynapSimCore object construction using higher level representaitons of the currents like gain ratio or time constant whenever applicable.

Parameters:

Idc (FloatVector, optional) – Constant DC current injected to membrane in Amperes, defaults to default_currents[“Idc”]
If_nmda (FloatVector, optional) – NMDA gate soft cut-off current setting the NMDA gating voltage in Amperes, defaults to default_currents[“If_nmda”]
r_gain_ahp (FloatVector, optional) – spike frequency adaptation block gain ratio, defaults to default_gain_ratios[“r_gain_ahp”]
r_gain_ampa (FloatVector, optional) – xcitatory AMPA synpse gain ratio, defaults to default_gain_ratios[“r_gain_ampa”]
r_gain_gaba (FloatVector, optional) – inhibitory GABA synpse gain ratio, defaults to default_gain_ratios[“r_gain_gaba”]
r_gain_nmda (FloatVector, optional) – excitatory NMDA synpse gain ratio, defaults to default_gain_ratios[“r_gain_nmda”]
r_gain_shunt (FloatVector, optional) – inhibitory SHUNT synpse gain ratio, defaults to default_gain_ratios[“r_gain_shunt”]
r_gain_mem (FloatVector, optional) – neuron membrane gain ratio, defaults to default_gain_ratios[“r_gain_mem”]
t_pulse_ahp (FloatVector, optional) – the spike pulse width for spike frequency adaptation circuit in seconds, defaults to default_time_constants[“t_pulse_ahp”]
t_pulse (FloatVector, optional) – the spike pulse width for neuron membrane in seconds, defaults to default_time_constants[“t_pulse”]
t_ref (FloatVector, optional) – refractory period of the neurons in seconds, defaults to default_time_constants[“t_ref”]
Ispkthr (FloatVector, optional) – spiking threshold current, neuron spikes if \(I_{mem} > I_{spkthr}\) in Amperes, defaults to default_currents[“Ispkthr”]
tau_ahp (FloatVector, optional) – Spike frequency leakage time constant in seconds, defaults to default_time_constants[“tau_ahp”]
tau_ampa (FloatVector, optional) – AMPA synapse leakage time constant in seconds, defaults to default_time_constants[“tau_ampa”]
tau_gaba (FloatVector, optional) – GABA synapse leakage time constant in seconds, defaults to default_time_constants[“tau_gaba”]
tau_nmda (FloatVector, optional) – NMDA synapse leakage time constant in seconds, defaults to default_time_constants[“tau_nmda”]
tau_shunt (FloatVector, optional) – SHUNT synapse leakage time constant in seconds, defaults to default_time_constants[“tau_shunt”]
tau_mem (FloatVector, optional) – Neuron membrane leakage time constant in seconds, defaults to default_time_constants[“tau_mem”]
Iw_0 (FloatVector, optional) – weight bit 0 current of the neurons of the core in Amperes, defaults to default_weights[“Iw_0”]
Iw_1 (FloatVector, optional) – weight bit 1 current of the neurons of the core in Amperes, defaults to default_weights[“Iw_1”]
Iw_2 (FloatVector, optional) – weight bit 2 current of the neurons of the core in Amperes, defaults to default_weights[“Iw_2”]
Iw_3 (FloatVector, optional) – weight bit 3 current of the neurons of the core in Amperes, defaults to default_weights[“Iw_3”]
Iw_ahp (FloatVector, optional) – spike frequency adaptation weight current of the neurons of the core in Amperes, defaults to default_currents[“Iw_ahp”]
C_ahp (FloatVector, optional) – AHP synapse capacitance in Farads, defaults to default_layout[“C_ahp”]
C_ampa (FloatVector, optional) – AMPA synapse capacitance in Farads, defaults to default_layout[“C_ampa”]
C_gaba (FloatVector, optional) – GABA synapse capacitance in Farads, defaults to default_layout[“C_gaba”]
C_nmda (FloatVector, optional) – NMDA synapse capacitance in Farads, defaults to default_layout[“C_nmda”]
C_pulse_ahp (FloatVector, optional) – spike frequency adaptation circuit pulse-width creation sub-circuit capacitance in Farads, defaults to default_layout[“C_pulse_ahp”]
C_pulse (FloatVector, optional) – pulse-width creation sub-circuit capacitance in Farads, defaults to default_layout[“C_pulse”]
C_ref (FloatVector, optional) – refractory period sub-circuit capacitance in Farads, defaults to default_layout[“C_ref”]
C_shunt (FloatVector, optional) – SHUNT synapse capacitance in Farads, defaults to default_layout[“C_shunt”]
C_mem (FloatVector, optional) – neuron membrane capacitance in Farads, defaults to default_layout[“C_mem”]
Io (FloatVector, optional) – Dark current in Amperes that flows through the transistors even at the idle state, defaults to default_layout[“Io”]
kappa_n (FloatVector, optional) – Subthreshold slope factor (n-type transistor), defaults to default_layout[“kappa_n”]
kappa_p (FloatVector, optional) – Subthreshold slope factor (p-type transistor), defaults to default_layout[“kappa_p”]
Ut (FloatVector, optional) – Thermal voltage in Volts, defaults to default_layout[“Ut”]
Vth (FloatVector, optional) – The cut-off Vgs potential of the transistors in Volts (not type specific), defaults to default_layout[“Vth”]

Returns:

DynapSimCore object instance

Return type:

DynapSimCore

property gain: DynapSimGain

Igain_ahp, Igain_ampa, Igain_gaba, Igain_nmda, Igain_shunt, Igain_mem

Type:: gain creates the high level gain ratios set by currents

get_full(size: int) → Dict[str, ndarray]

get_full creates a dictionary with respect to the object, with arrays of current values

Parameters:: size (int) – the lengths of the current arrays
Returns:: the object dictionary with current arrays given the size
Return type:: Dict[str, np.ndarray]

kappa_n: float | ndarray | Tensor | array = 0.75: Subthreshold slope factor (n-type transistor)

kappa_p: float | ndarray | Tensor | array = 0.66: Subthreshold slope factor (p-type transistor)

property layout: DynapSimLayout: layout returns a subset of object which belongs to DynapSimLayout

property time: DynapSimTime: time creates the high level time constants set by currents Ipulse_ahp, Ipulse, Iref, Itau_ahp, Itau_ampa, Itau_gaba, Itau_nmda, Itau_shunt, Itau_mem

update(attr: str, value: Any) → DynapSimCore[source]

update_current updates an attribute and returns a new object, does not change the original object.

Parameters:

attr (str) – any attribute that belongs to DynapSimCore object
value (Any) – the new value to set

Returns:

updated DynapSimCore object

Return type:

DynapSimCore

update_gain_ratio(attr: str, value: Any) → DynapSimCore[source]

update_gain_ratio updates currents setting gain ratio (Igain/Itau) attributes

Parameters:

attr (str) – any attribute that belongs to any DynapSimGain object
value (Any) – the new value to set

Returns:

updated DynapSimCore object

Return type:

DynapSimCore

update_time_constant(attr: str, value: Any) → DynapSimCore[source]

update_time_constant updates currents setting time constant attributes

Parameters:

attr (str) – any attribute that belongs to any DynapSimTime object
value (Any) – the new value to set

Returns:

updated DynapSimCore object

Return type:

DynapSimCore

property weight_bits: DynapSimWeightBits: weight_bits returns a subset of object which belongs to DynapSimWeightBits