Workshop 1 Model Risk

Workshop 1

Workshop 1: Model Risk - November 7

Managing machine learning model risk

Led by:

Agus Sudjianto, Head of Corporate Model Risk, WELLS FARGO

Jie Chen, Head of Statistics and Machine Learning, WELLS FARGO

An additional guest speaker:

Bernhard Hientzsch, Head of Model Library and Testing Development, WELLS FARGO

8:30 Registration and breakfast

9:00 Introduction

  • Machine learning applications in banking
  • Introduction to machine learning methodology
  • Trees and Random Forest Gradient Boosting Machine
  • Neural Network: Feed-Forward, Recurrent/LSTM, Convolutional Network
  • Generative Adversarial Network

10:30 Morning coffee break

11:00 Explainability and interpretation techniques

  • Understanding conceptual soundness of machine learning model
  • Global variable importance
  • Variable importance
  • Partial dependence plot and individual conditional expectation
  • Variance and derivative-based sensitivity
  • Local variable importance
  • Locally Interpretable Model and Effect (LIME, K-LIME, and LIME-SUP)
  • Total Derivative Effects

12:30 Lunch

1:30 Model risk  and validation of machine learning models

  • Data bias and mitigation
  • Conceptual soundness and explainability of machine learning
  • Replicability and benchmarking
  • Retraining and model monitoring
  • Implementation control

3:00 Afternoon coffee break

3:30 Model benchmarking and case study

  • Solving derivative pricing SDE Using Deep Learning
  • AI to Explain AI: Explainable Neural Networks

Guest Speaker:

Bernhard Hientzsch, Managing Director, ‎Head of Model, Library and Tools Development, WELLS FARGO