Tutorials

We wrote some notebooks to illustrate the functionalities of our algorithm:

1. In the SpQR tutorial we demonstrate the functionality of our NeuralSplineFlow bijector to evaluate and then sample feature distributions.

2. In the diglm tutorial we train a DIGLM model on a toy dataset with a [0, 1] label of the feature vectors. We demonstrate the capability of the model to be trained both for solve the logistic regression problem and the evaluation of feature distribution.

A more complicated application

We trained the DIGLM model on a Monte Carlo dataset simulating Beyond Standard Model events for a heavy Higgs boson.

We hope you enjoy the Higgs notebook.

Notebooks list

• Neural Spline Flow Tutorial
  • 1. Introduction
    • Package required to run this notebook
  • 2. Toy dataset
    • Plot helper function
  • 3. Initialization of NSF
  • 4. Model training
  • 5. Results …
  • … and conclusions
• DIGLM Tutorial
  • 1. Introduction
  • 2. Toy dataset
  • 3. Utilities: transforming datasets, sampling of distributions and plotting
  • 4. Building the DIGLM
  • 5. Training
  • 6. Results
    • About \(\lambda\)
• DIGLM on Higgs dataset
  • Package needed
  • Downloading data
  • Plotting features of the dataset
  • Slicing data: train, validation, test
  • Building the DIGLM model
  • Training
  • Results