Glossary of methods¶

This glossary defines the package-level terms used by jaxfne docs, examples, tests, and manifests.

Objective grammar¶

Emitter -> Source -> FieldProxy -> Probe -> Objective -> Optimizer -> Manifest

Term	Meaning in `jaxfne`
`Emitter`	Local state generator for voltage, spikes, synaptic traces, or typed neural state.
`Source`	Declared source tensor derived from emitter state under one source mode per run.
`FieldProxy`	Fixed computational projection from source tensors to named proxy readouts.
`Probe`	Selector and sampling operator for a readout family and channel layout.
`Objective`	Metrics, gates, nulls, and scores computed from readouts.
`Optimizer`	Bounded parameter search or gradient path over declared objectives.
`Manifest`	Strict JSON receipt for runtime, artifacts, hashes, truth gates, and validation status.

y_{n+1} = y_n + \Delta t\, f(y_n, t_n)

jaxfne uses explicit state updates for emitter dynamics and wraps hot time loops with jax.lax.scan where practical.

\Phi_{k,c} = \sum_n S_{k,n} W_{c,n}

S is the source tensor, W is a declared projection kernel, k indexes time or sample, c indexes channels, and n indexes source support.

\mathrm{lfp\_proxy}_{k,c} = \sum_n S_{k,n} K_{c,n}

The kernel K encodes the laminar or spatial projection used by the run.

\mathrm{csd\_proxy}_{k,c} = D_{zz}\Phi_{k,c}

D_{zz} is the declared depth-axis finite-difference operator.

\mathrm{eeg\_proxy}_{k,c} = \sum_n S_{k,n} L^{\mathrm{eeg}}_{c,n}

\mathrm{meg\_proxy}_{k,c} = \sum_n S_{k,n} L^{\mathrm{meg}}_{c,n}

L is the fixed readout kernel declared for the run.

Method	Package behavior
`jax.numpy` kernels	numerical arrays and vectorized tensor operations
explicit PRNG keys	deterministic seeded simulations and search paths
`jax.lax.scan`	hot time loops for emitter and trace updates
`jax.vmap`	candidate, seed, readout, and batch axes where supported
`jax.jit`	pure numerical kernels with I/O and plotting kept outside compiled functions
dtype policy	`float32` default with `float64` opt-in through JAX x64 configuration

Optimizer	Role
`SDR`	spectral descent response transform
`GSDR`	generalized spectral descent response transform
`AGSDR`	adaptive generalized spectral descent response path
random search	bounded stochastic search over declared parameter spaces

Object	Role
`Config` / `Configuration`	typed circuit, runtime, source, probe, and task specification
`Model` / `Net`	constructed computational graph and identity map
`Signals`	tensor output container with semantic selectors
`ProbeReport`	readout metadata and interpretation status
`RuntimeReport`	backend, dtype, device, and timing receipt