Methodology

Short-term goals

• PDF parametrization
  • Fit preprocessing
  • Neural network architecture:
    • Single-layer for each PDF
    • One single (multilayer or perhaps deep) NN for all PDFs
• Minimization algorithms
  • CMA algorithm: validation
    • Closure test
    • Check against reweighting
    • Check of positivity
  • CMA algorithm: optimization
    • Grid search

 

---

 

Mid- to long-term goals

• Minimization algorithms
  • CMA algorithm: optimization
    • Gradient-based methods
  • Weight minimization + other new methods