How adamjamessheppard/lorenz-training-set works
This is a numerical-research solver for the forecast step of a continuous-discrete Bayesian filter on stochastic Lorenz-63, with sequential data-assimilation and dataset-generation capabilities. The main workflow is truth simulation and observations, density forecast, Bayesian analysis, then feeding the posterior into the next forecast; accepted forecast pairs can be exported into trajectory-safe datasets. The graph also shows positivity treatment and validation. Some solver internals and experiment wiring are unsampled, so those links are omitted unless documented in the README or shown in excerpts.
Main entry points: dataset.py Trajectory dataset · core.py DG forecast solver · validation.py Analytic and MC validation · solver.py Forecast entry point
Components 14
Model and states
- Truth simulator
lorenz_fpe/core.py - Observation model
lorenz_fpe/core.py - Density state
lorenz_fpe/core.py
Forecast numerics
- DG forecast solver
lorenz_fpe/core.py - Positivity limiter
lorenz_fpe/afc.py - Local QP projection
lorenz_fpe/local_projection.py - AFC update
lorenz_fpe/afc.py - Forecast entry point
solver.py
Data assimilation
- Bayesian analysis
lorenz_fpe/core.py
Dataset and validation
- Trajectory dataset
lorenz_fpe/dataset.py - Dataset entry point
generate_dataset.py - Analytic and MC validation
lorenz_fpe/validation.py - Finite-volume reference
lorenz_fpe/finite_volume.py
Other components
- Research user
Connections 16
- Research user → Forecast entry point: runs forecast
README.md - Forecast entry point → DG forecast solver: invokes solver
solver.py - DG forecast solver → Density state: advances density
README.md - DG forecast solver → Positivity limiter: limits density
README.md - Research user → Dataset entry point: starts generation
README.md - Dataset entry point → Trajectory dataset: creates dataset
generate_dataset.py - Trajectory dataset → Truth simulator: simulates truth
lorenz_fpe/dataset.py - Trajectory dataset → Observation model: configures observations
lorenz_fpe/dataset.py - Trajectory dataset → DG forecast solver: forecasts posterior
README.md - Trajectory dataset → Bayesian analysis: updates forecast
README.md - Bayesian analysis → Density state: returns posterior
README.md - DG forecast solver → Local QP projection: projects Q2 state
lorenz_fpe/local_projection.py - AFC update → Local QP projection: uses local projection
lorenz_fpe/afc.py - Analytic and MC validation → DG forecast solver: validates solver
lorenz_fpe/validation.py - Analytic and MC validation → Finite-volume reference: compares reference
lorenz_fpe/validation.py - Trajectory dataset → Density state: exports tensors
README.md