Skip Navigation
Search

AMS 520, Machine Learning in Quantitative Finance 

Description:

This course will merge ML and traditional quantitative finance techniques employed at investment banks, asset management, and securities trading firms. It will provide a systematic introduction to statistical learning and machine learning methods applied in Quantitative Finance. The topics discussed in the course fall broadly into four categories which (as time permits) will be discussed in this order:

1) Probabilistic Modeling:  Bayesian vs. frequentist estimation, bias-variance tradeoff, sequential Bayesian updates, model selection and model averaging; Probabilistic graphical models and mixture models; Multiplicative Weights Update Method Bayesian regression and Gaussian processes.

2) Feedforward neural networks: Feedforward architecture; Stochastic gradient descent and backpropagation algorithm; Non-Linear Factor Modeling and applications in asset pricing; Convolutional neural networks; Autoencoders.

3) Sequential Learning: Linear time series models; Probabilistic sequence modeling – Hidden Markov Models and particle filtering; Recurrent Neural Networks; Applications in finance.

4) Reinforcement Learning: Markov decision process and dynamic programming methods (Bellman equations and Bellman optimality); Reinforcement learning methods (Monte-Carlo methods, policy-based learning, TD-learning, SARSA, and Q-learning); Deep reinforcement learning; Applications of reinforcement learning in finance.
 
 
Prerequisites:   AMS 572 and AMS 595; or AMS 561; or based on knowledge of Python per instructor's consent
 

Offered in fall semester; 3 credits; ABCF grading

 

Course Materials:  

Required:

"Machine Learning in Finance: From Theory to Practice" by Paul Bilokon, Matthew F. Cixon, and Igor Halperin; 2020, Springer Publishing; ISBN: 978-3030410674 (eBook rental)

 

Recommended Additional Readings:

"Machine Learning for Algorithmic Trading" by Stefan Jansen; 2nd edition, 2020, Packt Publishing; ISBN: 978-1-83921-771-5

"The Elements of Statistical Learning: Data Mining, Inference, and Prediction" by Trevor Hastie, Robert Tibshirani, and Jerome Friedman (available at http://statweb.stanford.edu/~tibs/ElemStatLearn/ )

"Machine Learning for Factor Investing" by Guillaume Coqueret and Tony Gida, R Version, CRC Press (available at:  http://www.mlfactor.com/)

"Reinforcement Learning: An Introduction" by Richard S. Sutton and Andrew G. Barto, 2nd edition, 1998, The MIT Press; ISBN: 978-0262193986

"Python for Finance: Analyze Big Finanical Data" by Yves Hilpish, 1st edition; ISBN: 978-1491945285

 

 

Learning Outcomes:

Demonstrate understanding of probabilistic modeling:

  • Bayesian vs. frequentist estimation, bias-variance tradeoff, sequential Bayesian updates, model selection and model averaging;
  • Probabilistic graphical models and mixture models;
  • Multiplicative Weights Update Method
  • Bayesian regression and Gaussian processes.

Demonstrate understanding of feedforward neural networks:

  • Feedforward architecture;
  • Stochastic gradient descent and backpropagation algorithm;
  • Non-Linear Factor Modeling and applications in asset pricing;
  • Convolutional neural networks.

Demonstrate understanding of methods for sequential learning:

  • Linear time series models;
  • Probabilistic sequence modeling – Hidden Markov Models and particle filtering;
  • Recurrent Neural Networks;
  • Applications in finance.

Demonstrate understanding of dynamic programming and reinforcement learning algorithms:

  • Markov decision process and dynamic programming methods (Bellman equations and Bellman optimality);
  • Reinforcement learning methods (Monte-Carlo methods, policy-based learning, TD-learning, SARSA and Q-learning);
  • Deep reinforcement learning;
  • Applications of reinforcement learning in finance.