BachelorThesis

A Thesis Submitted to EEMCS Faculty Delft University of Technology, in Partial Fulfilment of the Bachelor of Computer Science and Engineering Requirements. Code for bachelor thesis. This is for Research Project for the year 2024 Q2.

How to run

We use a conda environment to run the code. It is running on WLS 2 with Ubuntu.

To use conda in WSL, you need to set up WSL first, then install conda in WSL. Then you can create a conda environment and install the packages required.

After that, you can run the code in the environment.

First, run the LCDB_loclised.py to get the curves required.

Afterwards, you can run exp_1.ipynb to check metrics, exp_2.ipynb to see where the LC-PFN model suffers and training_lcdb.ipynb to train the model(currently causing the trained model to not give proper results leading a straight-line at 0.8).

If you have issues with the LC-PFN, change all the from lcpfn.lcpfn import * to from lcpfn import * in the folder lcpfn.

Additional information:

For experiment 2, you need the baseline (Last 1 and mmf4). You can get it from the LCDB Github repo follow the instructions to get the "df_total.gz" file. Then you can run the experiment 2 notebook. If you run it with getting the baseline you will only get the MSE for LC-PFN.

Code for LC-PFN from :

https://github.com/automl/lcpfn

@inproceedings{
anonymous2023efficient,
title={Efficient Bayesian Learning Curve Extrapolation using Prior-Data Fitted Networks},
author={Adriaensen, Steven and Rakotoarison, Herilalaina and Müller, Samuel and Hutter, Frank},
booktitle={Thirty-seventh Conference on Neural Information Processing Systems},
year={2023},
url={https://openreview.net/forum?id=xgTV6rmH6n}
}

Code for LCDB from : https://github.com/fmohr/lcdb/

@inproceedings{lcdb,
  title={LCDB 1.0: An Extensive Learning Curves Database for Classification Tasks},
  author={Mohr, Felix and Viering, Tom J and Loog, Marco and van Rijn, Jan N},
  booktitle = {Machine Learning and Knowledge Discovery in Databases. Research Track - European Conference, {ECML} {PKDD} 2022, Grenoble, France, September 19-24, 2022},
  year={2022}
}