Chapter 23: Robust and Non-Parametric Estimation

Learning Objectives

• Construct Huber M-estimators and read their behaviour off the influence function
• Compute the breakdown point of an estimator and use it to compare robustness
• Derive the Nadaraya–Watson kernel regressor and understand bias–variance trade-offs in bandwidth selection
• Characterize Gaussian process regression as Bayesian non-parametric inference and interpret its predictive uncertainty
• Formulate an estimation problem as an end-to-end learning task and compare it to model-based baselines
• Explain deep unfolding as a hybrid between iterative algorithms and neural networks