This course aims at introducing the fundamental concepts of Reinforcement Learning (RL), and develop use cases for applications of RL for option valuation, trading, and asset management.
By the end of this course, students will be able to
- Use reinforcement learning to solve classical problems of Finance such as portfolio optimization, optimal trading, and option pricing and risk management.
- Practice on valuable examples such as famous Q-learning using financial problems.
- Apply their knowledge acquired in the course to a simple model for market dynamics that is obtained using reinforcement learning as the course project.
Prerequisites are the courses "Guided Tour of Machine Learning in Finance" and "Fundamentals of Machine Learning in Finance". Students are expected to know the lognormal process and how it can be simulated. Knowledge of option pricing is not assumed but desirable.
MDP and Reinforcement Learning
MDP model for option pricing: Dynamic Programming Approach
MDP model for option pricing - Reinforcement Learning approach
Pablo Torre completed this course, spending 12 hours a week on it and found the course difficulty to be very hard.
The material covered in the course is mostly focused around the mathematics of the Bellman and Black Scholes equations. The professor takes several weeks attempting to relate between the two in order to provide a mathematical framework to price options...
The material covered in the course is mostly focused around the mathematics of the Bellman and Black Scholes equations. The professor takes several weeks attempting to relate between the two in order to provide a mathematical framework to price options through reinforcement learning, but fails to present this information in a way that is straight forward to put into practice. The problem is further complicated by inconsistent use of notation and excessive amount of details which are not needed to put the concepts into code.
The programming assignments are ridden with problems, from graders that only accept an answer which is technically wrong and must be entered with a "known bug" , to assignment instructions which include many complicated formulas... but not the ones needed to complete the assignment. On top of this the assignments are implemented in a hacky way that doesn't make full use of Coursera's auto-grader and instead requires copying hash-key's back and forth in order to submit answers, or even building your own local environment to execute the assignment notebooks as the cloud version gets stuck. The final peer graded assignment is specially bad, as the problems above get compounded by the difficulties of finding peers to review your assignment on a sparsely taken course.
The discussion forums were abandoned by course staff and the only support available comes from other students who have "figured it out" on their own, or had posted questions showing their code seeking help on messages which have been reported for months without a solution.
Albert Moon is taking this course right now, spending 5 hours a week on it and found the course difficulty to be medium.
A very poor class: the material dive deep in math but with no explanation and moreover the material does not help to solve the homework or to reach the course goals.
The forum is almost dead.
The only acceptable points are the jupyter notebooks wich are interesting to solve (but you will need to find the knowledge outside of the course as I mentioned before).