BrainPy-style Modeling Guide

This guide provides an in-depth tour of the BrainPy-style modeling API — the high-level, composable neuron, synapse, projection, and readout building blocks shipped under brainpy.state. The BrainPy-style API is the recommended entry point for building and training spiking neural networks with this library.

Note

This guide covers the BrainPy-style API only. For NEST-compatible models (iaf_psc_alpha, aeif_cond_exp, STDP synapses, etc.), see the NEST-Compatible Models — Status & Limitations page — those models are under active development and have their own usage notes.

What this guide covers

• Architecture: How brainpy.state composes Dynamics, Neuron, Synapse, projections, and readouts on top of brainstate’s state management.

• Neurons: Building blocks for spiking computation — LIF, ALIF, AdEx, HH, Izhikevich, and their refractory variants — and how state variables are represented.

• Synapses: Exponential, alpha, AMPA, GABA, and NMDA dynamics; how conductance-based (COBA), current-based (CUBA), and magnesium-block outputs attach to a postsynaptic population.

• Projections: Network-level wiring — AlignPostProj, delta projections, and short-term plasticity hooks — for connecting populations.

Why these concepts matter

Working through this guide will help you:

• Design complex spiking networks with clean, composable code.

• Use surrogate gradients to train BrainPy-style neurons end-to-end.

• Manage simulation state and physical units (saiunit) idiomatically.

• Integrate cleanly with the rest of the BrainX ecosystem.

The pages build on one another, so we recommend reading them in order if you’re new to brainpy.state.