CAML

Source code for the paper "A Few Models to Rule Them All: Aggregating Machine Learning Models" at LWDA 2023.

Installation

poetry install
poetry shell

Getting Started

See demo.py for a simple example.

This implementation of CAML currently supports time series forecasting models only provided by the Darts library. However, it can easily adapted for other libraries and model types (e.g. sklearn).

Adding model architectures

To add a new model architecture, you need to create a new class that inherits from ObjectiveFactory and implements the create and build_model methods. See also the prebuilt factories for examples.

Citation

If this is useful for your research, please consider citing the paper (PDF):

@inproceedings{Siepe2023,
    author = {Florian Siepe and Phillip Wenig and Thorsten Papenbrock},
    title = {A Few Models to Rule Them All: Aggregating Machine Learning Models},
    booktitle = {Proceedings of the Conference Lernen, Wissen, Daten, Analysen},
    numpages = {12},
    year = {2023},
    series = {LWDA '23},
    location = {Marburg, Germany},
    publisher = {CEUR Workshop Proceedings},
    url = {https://ceur-ws.org/XXXX}
}