This section of the ML-INFN Confluence Space contains the Knowledge Base of fully implemented use cases. This has been created in order to provide new users getting close to Machine learning with concrete examples, with step by step guides for reproducibility.

The division into categories is multidimensional 

  • Dimension 1: per goal (classification, regression, clustering, technological ...)
  • Dimension 2: per Machine Learning algorithms (CNN, Auto encoders, LSTM, GraphNet, ...)
  • Dimension 3: per scientific field (High Energy Physics, Gravitational Waves, Medical Physics, ...) 
  • Dimension 4: per type of used tool

and is implemented via Confluence labels.

Table of Use cases

Name and LinkGoalML AlgorithmsScientific FieldML ToolsComments
Btagging in CMS (templated version)ClassificationCNN, LSTMHigh Energy PhysicsKeras + TensorflowRealistic application
LHCb Masterclass, with KerasDensity estimation and classificationMLPHigh Energy PhysicsROOT + Keras + TFIntroductory tutorial
MNIST in a C headerClassificationMLP
KerasFree-styling tutorial

LUMIN: Lumin Unifies Many Improvements for Networks


CNN, RNN, GNNHigh Energy PhysicsPyTorchPackage use examples
INFERNO: Inference-Aware Neural OptimisationClassificationNNHigh Energy PhysicsKeras + TensorflowTechnique application example
An introduction to classification with CMS dataClassificationFisher, BDT, MLPHigh Energy PhysicsScikit-learn, TF2

Tutorials for Master Students

Virgo Autoencoder tutorialData CompressionAutoencoderGeneral RelativityPython KerasTutorial for student
Distributed training of neural networks with Apache SparkTechnologicalDNNHigh Energy PhysicsSpark + BigDLTutorial
FTS log analysis with NLPSelf-supervised, clusteringNLPHigh Energy Physics, ComputingWord2Vec + Rake + sklearnTutorial

Image Inpainting tutorial: how to digitally restore damaged images


CNN U-NetApplied PhysicsKeras + Sci-kit image, PIL, OpenCV, matplotlibTutorial

Signal/background discrimination for the VBF Higgs four lepton decay channel with the CMS experiment using Machine Learning classification techniques


ANNs, RFHigh Energy PhysicsKeras , TensorFlow, Scikit-learnTutorial

Explainability of a CNN classifier for breast density assessment

Explainability AI

CNNMedical PhysicsKeras, TensorflowTutorial

ML for smart caching


ML/RLHigh Energy Physics, Computing, CacheKeras, Tensorflow, sklearnDemo, playground

Signal-background Classification with Parametric Neural Networks


pNNHigh Energy PhysicsKeras + TensorFlow 2Tutorial
New Physics Learning MachineDensity EstimationNNHigh Energy PhysicsKeras, Tensorflowpackage + tutorial
MLaaS4HEP for the Higgs boson ML challengeTechnologicalDT, MLPHigh Energy PhysicsXGBoost, Keras + TensorFlow 2, PyTorchTutorial
Object counting with c-ResUnetSupervised learning, Semantic SegmentationCNN, ResUnetComputer VisionPyTorch, fastaiReal application, Tutorial
Fast classifier-based goodness of fit test for online data quality monitoringDensity estimationLogistic Regression, Kernel MethodsHigh Energy PhysicsPytorch, Falkon libraryTutorial + Realistic application

How to insert a new use case

Follow the instructions provided in the How To: Create a KB entry

Once you finish with the creation of the page don't forget to edit the page "Machine Learning Knowledge Base" (this same page!) and add the use case in the "Table of Use cases". 

  • No labels