Menu
Guidelines for developing NeuroJAX (OSL-JAX) components.
Installer avec Codex ou Claude Copiez ce prompt, collez-le dans Codex, Claude ou un autre assistant, puis laissez-le vérifier la page du skill et l'installer pour vous.
Basé sur la classification professionnelle SOC
Validate BEM Forward Model Accuracy using the MNE CTF Phantom dataset.
Validate Inverse Solver Accuracy using the MNE Elekta Phantom dataset.
Quantify Inverse Solver Leakage and Resolution using a Synthetic Phantom.
Detect Loss of Consciousness (LOC) using SINDy-based bifurcation analysis on EEG.
Validate Artifact Subspace Reconstruction (ASR) using the SSVEP with Artifact Trials dataset (ds004745).
Guidelines for implementing biophysical neural mass models in NeuroJAX.
| name | neurojax_dev |
| description | Guidelines for developing NeuroJAX (OSL-JAX) components. |
NeuroJAX follows the "Kidger Stack" philosophy:
equinox.Module. State is explicit and immutable.lineax for linear solves and optimistix for non-linear optimization.numpy (except for I/O); use jax.numpy.jaxtyping to enforce shapes, e.g., Float[Array, "time sensors"].src/neurojax/glm.py: Mass-univariate statistics.src/neurojax/inverse/: Beamformers and source reconstruction.src/neurojax/models/: Biophysical and generative models (Diffrax, Equinox).src/neurojax/utils/: Bridges and helpers.When solving $Ax=b$, prefer lineax over jnp.linalg.solve:
operator = lx.MatrixLinearOperator(A)
solution = lx.linear_solve(operator, b, solver=lx.QR())
Passing key is mandatory for stochastic operations. Split keys early:
keys = jax.random.split(key, num=100)
jax.vmap(func)(keys)