Guides¶
This section contains the hands-on documentation for neuraLQX. The goal of these pages is to help you use the library effectively by understanding the core abstractions, the data model, and the workflows you will use in real experiments.
If you are new to the package, the best reading order is:
Graphs: what the kinematics live on (vertices/edges, minimal loops, embeddings).
Hilbert spaces: what a configuration is and how degrees of freedom are encoded.
Gauge groups: how Gauß constraints and gauge copies are represented and evaluated.
LQX models: how physical operators and constraints are built from the kinematics.
Solver: the end-to-end VMC workflow, logging, checkpoints, and reproducibility.
You can jump directly to any guide below.
Note
These guides are written as API-facing documentation, not as a paper. They are meant to be read while you implement models, run sweeps, and debug training runs.
Quick links¶
The sections below provide direct links to the most commonly referenced guides.
Graph construction, multiedges, embeddings, and minimal loops. Use this when defining kinematics or building loop-based operators.
Configuration encoding, cutoffs, gauge copies, and constrained (gauge-invariant) representations.
Gauge-group descriptors and Gauß constraint operators (local, computational, and JAX backends).
Model interfaces, constraints, and the Euclidean / spherical / QRLG families.
End-to-end optimisation pipeline: samplers, networks, schedules, checkpoints, MPI, and analysis.