Tutorial

This guide demonstrates how to use PyEPO for modeling and solving predict-then-optimize problems.

• Model
  • User-defined Models
  • Pre-defined Models
• Data
  • Data Generator
  • optDataset
  • optDatasetKNN
• Two-Stage Method
• Auto Grad Functions
  • Smart Predict-then-Optimize Loss+ (SPO+)
  • Differentiable Black-box Optimizer (DBB)
  • Negative Identity Backpropagation (NID)
  • Differentiable Perturbed Optimizer (DPO)
  • Perturbed Fenchel-Young Loss (PFYL)
  • Implicit Maximum Likelihood Estimator (I-MLE)
  • Adaptive Implicit Maximum Likelihood Estimator (AI-MLE)
  • Noise Contrastive Estimation (NCE)
  • Contrastive Maximum A Posteriori Estimation (CMAP)
  • Learning to Rank (LTR)
  • Perturbation Gradient (PG)
  • Parallel Computation
• Solution Pool
• Training
  • Common Setup
  • Training with SPO+
  • Training with DBB
  • Training with NID
  • Training with DPO
  • Training with PFYL
  • Training with I-MLE
  • Training with AI-MLE
  • Training with NCE
  • Training with LTR
  • Training with PG
• Evaluation
  • Regret
  • Unambiguous Regret