Work program

- Tools to learn:
- `git`, `uv`, `scikit-learn`, s3, MLFlow, Argo, `quarto`.
- scikit-learn: Pipeline, transformer, estimator, CV, transform/fit/predict.
- How to define and use a configuration on SSP Cloud.
- Regarding machine learning algorithms:
- Read the introduction to [ensemble methods](https://inseefrlab.github.io/DT_methodes_ensemblistes/);
- Find and read papers on spatial interpolation using tree-based methods, and produce a literature review.
- Establish the list of algorithms to compare.
- Define the list of metrics to compare.
- Build the datasets:
- Define what sort of data we want to work on.
- Gather real datasets;
- Generate artificial datasets (using `gstools`?);
- Build code that automatically trains a suite of algorithms;
- Train basic models;
- Tune hyperparameters;
- Log models using MLFlow;
- Orchestrate training using Argo (optional).
- Prepare an automated report using Quarto.

Outputs:

- A fully reproducible evaluation pipeline in a clean Github repository.
- A report describing the whole project.
- A slideshow presenting the results.
- A series of clean and well-documented datasets usable for other benchmarks.

Start with the linked introduction to ensemble methods and review the repository's existing notebooks, then define the algorithms, metrics, datasets, and reproducibility approach. Done means a clean GitHub repository containing the automated evaluation pipeline, documented datasets, a Quarto report, and a slideshow presenting the results.

Scritto dal modello di indicizzazione a partire dal testo della issue.