Exp2Syn

class braincell.synapse.Exp2Syn(size, name=None, tau1=Quantity(0.1, 'ms'), tau2=Quantity(10., 'ms'), e=Quantity(0., 'mV'))

NEURON-compatible Exp2Syn template.

This class follows the dynamics in NEURON’s exp2syn.mod:

• states: A and B in uS

• decay: A' = -A / tau1, B' = -B / tau2

• conductance: g = B - A

• inward-positive point current: I = g * (e - V_post)

• step-boundary event: A <- A + weighted_pre_drive * factor and same for B

Parameters:

• size (int | Sequence[int] | integer | Sequence[integer]) – Packed synapse shape.

• name (str | None) – Runtime node name.

• tau1 (Array | ndarray | bool | number | bool | int | float | complex | Quantity | Callable) – Positive rise time constant, default 0.1 ms.

• tau2 (Array | ndarray | bool | number | bool | int | float | complex | Quantity | Callable) – Decay time constant greater than tau1, default 10 ms.

• e (Array | ndarray | bool | number | bool | int | float | complex | Quantity | Callable) – Reversal voltage, default 0 mV.

apply_events(payload, V_post=None)

Apply one already-aggregated event payload vector.

compute_derivative(V_post=None)

Compute the derivative of the channel’s state variables.

This method should be implemented by subclasses to calculate how the channel’s state changes over time.

Parameters:

• *args – Variable length argument list.

• **kwargs – Arbitrary keyword arguments.

Raises:

NotImplementedError – This method must be implemented by subclasses.

current(V_post)

Calculate the current for this ion channel.

This method should be implemented by subclasses to compute the current based on the channel’s specific properties and state.

Parameters:

• *args – Variable length argument list.

• **kwargs – Arbitrary keyword arguments.

Raises:

NotImplementedError – This method must be implemented by subclasses.

root_type

alias of HHTypedNeuron

classmethod validate_parameter_values(parameters)

Require tau1 < tau2 elementwise.

Return type:

None

Notes

The comparison runs in numpy, not u.math/jnp. Under an open JAX trace every jax operation returns a tracer even when all its inputs are concrete, so bool() on a u.math.any reduction would raise TracerBoolConversionError and make this class impossible to construct inside a jitted setup function. numpy stages nothing, and it is what the sibling positive() validator already uses.

Both mantissas are taken in ms rather than compared as quantities, so that a caller mixing us and ms is ordered on physical duration rather than on raw magnitude.