CTLearn Manager documentation

The CTLearn Manager is a companion python package to the CTLearn deep learning IACT event reconstruction package. It is designed to guide users from creating a CTLearn model, training, testing, and DL2 analysis to benchmark the model. It also offers a variety of tools to compare models between each other as well as with the standard RF DL2 files. It was developed thanks to the stereo system of SST-1M and the LST-1. The basic principle of the Manager revolves around a ctlearn_model_index file, that stores all the relevant data for you, such as characteristics of your models, IRFs, range of validity etc. This enables for a user-friendly interface to IACT event reconstruction and benchmarking with CTLearn. A series of notebooks allows the user to go through the full range of functionalities of the manager.

GitHub repository

Note

It is highly recommended to follow the steps from the beginning in order for the manager to know as much as possible about your models, as it stores data during every step.

Contents:

• 1. Creating CTLearn Model Managers
  • Create a model manager
  • Visualisation
• 2. Training
• 3. Tri-Model Managers
• 4. Testing
  • Testing from scratch
  • Existing MC DL2 files (not recommended)
• 5. IRFs
  • Produce IRFs
  • Visualization
• 6. Monte-Carlo DL2 analysis
  • Applying cuts to the data
  • DL2 Visualisation
• 7. DL1 to DL2 on real data
  • Standard data
  • lstchain data
  • Tri-Model Collection
• 8. Real data DL2 analysis
  • DL2DataProcessor
  • RFCounterpart
  • Combinator2000
  • WhoIsBetter
• 9. Export and import curves
  • Export curves
  • Import curves

API Reference

model_manager	CTLearnModelManager Class.
tri_model	CTLearnTriModelManager class for handling three CTLearn models: direction, energy, and type.
tri_model_collection	Plotting and predicting with a collection of CTLearnTriModelManager models.
utils.DL2_processing
utils.RF_counterpart
utils.utils