Notebooks¶

This project follows a two-phase workflow, stage by stage: prototype in a notebook, then promote the validated logic into the dispcraft library (see the API Reference) once it's shown to work — after that, the notebook becomes a thin driver that imports from the library and shows results, rather than duplicating logic.

The notebooks below are rendered here exactly as executed (outputs kept, not re-run by this site), in the order the project's stages were done. Each one's own markdown cells carry the detailed narrative; Status_Report.md and CLAUDE.md in the repository summarize findings across notebooks.

Notebook	Stage
1-Intro_sujet	Stage 1 — physical instrument model
2-Intro_data	Stage 2 — ground-test data loading and exploration
3-Intro_ML	Stage 3 — per-dataset physical calibration
4.1-ML_Comparison	Stage 4 Phase 1 — ML residual model comparison
4.2-Per_Dataset_Pipeline	Stage 4 Phase 2 — per-dataset hybrid pipeline
4.3-Multi_Dataset_Fitting	Stage 4 Phase 3 — joint multi-dataset fit
4.5-Comparison_With_Published_Results	Stage 4 Phase 5 — literature comparison
5.1-PyTorch_Migration	Stage 5 Phase 1 — sklearn → PyTorch/Lightning
5.2-Field_Dependent_Parameters	Stage 5 Phase 2 — field-dependent correction tier
5.3-Zeroth_Order_Dispersion	Stage 5 Phase 3 — 0th-order chromatic dispersion
5.4-BGS_Model	Stage 5 Phase 4 — BGS (blue grism) calibration
6-Status_Report	Stage 6 — final results assembly
8-Chebyshev_Residual	Stage 8 — Chebyshev polynomial residual model

Use the sidebar to open any of them.