SPRACE ML-Physics Meeting

Tuesday 9 May 2017, 13:00 → 14:00 America/Sao_Paulo

13:00 → 14:00 Slides and Discussion

Speakers: Alexandre Alves (Universidade Federal de São Paulo / UNIFESP), Jose Cupertino Ruiz Vargas (UNESP - Universidade Estadual Paulista (BR)), Marco André Ferreira Dias (IFT-UNESP), Raphael Cóbe, Thiago Tomei Fernandez (UNESP - Universidade Estadual Paulista (BR)), Vitor Finotti Ferreira (UNESP - Universidade Estadual Paulista (BR))

Jet classification (Jose)

Jet classification (Rafael)

Rafael's presentation - basic neural network (625 -> 256 -> 256 -> 2)

Suggestions for next steps

1. Improve the neural network architecture 
2. Test other models of neural networks (like a convolution network)
3. Do cross-validation

Thiago suggestions:

• Run the basic network for different pT ranges
• Check out how to visualize the network (whatever that means)

---

Jose's presentation - Random Forest Classification

• Is there any way to tune the classifier to sit "at a given point of the ROC curve"?

---

Other comments:

• Let us try to implement a FLD, to follow on the paper
• Thiago has to implement the tau21 classifier