shml 0.1

lightweight pdata cleaning/processing/plotting/ML training library for use with an ATLAS BSM dihiggs search

shml: routines to automate machine learning experiments for a X -> SH -> bbyy search

This module aims to provide a set of functions that, when composed, can run a pipeline capable of:

• going from .root files to parquet files via uproot and awkward
• constructing useful kinematic quantites for training
• applying a chosen or manual preselection
• configuring any additional processing, e.g. weight normalization, feature scaling
• access event data that's prepared for pytorch using shml.torch_dataset.EventDataset

still to do:

• infra to run ml experiments in a GPU or CPU environment via pytorch-lightining

Usage

To see currently usable implemented functionality, check the examples folder.

Install

For preprocessing only:

python3 -m pip install shml

For ML extras (pytorch, plotting):

python3 -m pip install shml[ml]