Feed Forward Event-driven Neural Network (FFEvNN)

class FFEvNN(EvNN)

A feed‑forward EvNN with layered connectivity.

Constructor

FFEvNN(
    layers: list[int],
    in_size: int,
    neuron_model: NeuronModel,
    max_solver_time: float,
    key: jax.random.PRNGKey | None = None,
    init_delays: jnp.ndarray | float | int | None = None,
    dlim: float | None = None,
    buffer_capacity: int = 500,
    solver_stepsize: float = 1e-2,
    solver_checkpoints: int | None = None,
    max_event_steps: int = 100,
    output_no_spike_value: float | None = None,
    root_finder=None,
    stepsize_controller=None,
    solver=None,
    dtype=jnp.float32,
    **neuron_model_kwargs,
)

Behavior

• Builds a strict feed‑forward mask connecting layer ℓ to layer ℓ+1.
• Marks the first layer as input_neurons and the last as output_neurons.
• Delegates the rest to the base EvNN constructor.