46838 Machine Learning for Computational Finance
46838 Machine Learning for Computational Finance
Spring Mini 4, 1999
Prof. Roni Rosenfeld
School of Computer Science, Carnegie Mellon University
The purpose of this course is to give a broad introduction to the
techniques of machine learning, and to place those techniques within
the context of computational finance. Machine learning is concerned
with building computer programs that learn and improve with
experience. The class will start out with an introduction of the
underlying philosophy and methodology of machine learning, and then
move on to hands on application of such techniques as neural nets and
decision trees to real financial datasets.
Time and Place:
- Pittsburgh:
- section A: Mondays 1:00PM--4:00PM, FastLab, GSIA.
- section E: Mondays, 5:30PM--8:30PM, FastLab, GSIA.
- New York: Mondays 5:30PM--8:30PM
- London: Mondays 6:00PM--9:00PM (GMT)
Class runs from March 8th till April 26th inclusive. No class on March 22nd (CMU Spring break).
NB: The week of April 19th, the London and Pittsburgh A classes
will be taught on Wednesday, April 21st.
Instructor:
Roni Rosenfeld
(roni@cs.cmu.edu), Wean Hall 4109, 412/268-7678. Fax: 412/268-5576.
Office hours : Thursday, 11AM-noon, or by appointment (set
up via email).
Teaching Assistant:
Rosie Jones
(rosie+46838 @ cs.cmu.edu), Cyert Hall, 412/268-8492
Office hours (Recitation):
- Pittsburgh: Saturdays 10AM--NOON (EST)
- New York: Saturdays 11AM--NOON (EST)
- London: Saturdays 3PM--4PM (GMT)
These sessions will be broadcast to NY, London at
the times listed above.
Location: FASTLab, or by appointment
Saturday April 10th, office hours will be in NY
thus there will be no London office hours on
this date; alternative arrangements will be made
in class.
(no office hours March 20th)
Course Secretary:
Dorothy Zaborowski (daz+@cs.cmu.edu), Wean Hall 4116. 412/268-3779.
Administrative Help:
Pittsburgh: Norene Mears (nm10+@andrew.cmu.edu), GSIA 128, 412-268-7358
New York: Yuvelin Tejeda (tejeda+@andrew.cmu.edu), 212-584-0925
London: Mandy Hossami (Mandy.Hossami@dresdnerkb.com) 011-44-171-475-6187
Textbooks:
Machine Learning, by Tom Mitchell, McGraw Hill.
Available at the CMU book store, or can be ordered online.
Grading:
Will be based on homeworks (60%) and a final (40%).
Note: you must pass the final to pass the course.
Policy on homework:
- Unless otherwise stated, all homework assignments are to be done
completely on your own, with no communication with anyone except the
TA or the instructor. This policy will be strictly enforced.
Announcements Last Updated 4/22/99
Homework Assignments:
- Assignment 1 (due March 15. 1999 at the beginning of class) -- email your solutions by the due date (ASCII (i.e. save as text) preferred).
- Assignment 2 (due March 29. 1999 at the beginning of class) -- email your solutions by the due date (as text ONLY), or fax to Rosie Jones on +1 412 268-6298
- Assignment 3 (due April 5. 1999 at the beginning of class) -- email your solutions by the due date, or fax to Rosie Jones on +1 412 268-6298. You may submit your assignment solutions as Microsoft word documents.
- Assignment 4 (due April 12. 1999 at the beginning of class) -- email your solutions by the due date, or fax to Rosie Jones on +1 412 268-6298. You may submit your assignment solutions as text, postscript, or Microsoft word documents.
- Assignment 5 (due April 19. 1999 at the beginning of class) -- email your solutions by the due date, or fax to Rosie Jones on +1 412 268-6298. You may submit your assignment solutions as text, postscript, or Microsoft word documents, but only as a single file (zip files will not be opened).
Faxed submissions are preferred.
- Assignment 6 (due April 26. 1999 before the beginning of class) -- email your solutions by the due date, or fax to Rosie Jones on +1 412 268-6298. You may submit your assignment solutions as text, postscript, or Microsoft word documents, but only as a single file (zip files will not be opened).
Faxed submissions are preferred.
MATLAB Tutorial
- You can work through this brief tutorial on your own time. Some office-hour time may be devoted to MATLAB familiarization also.
Tentative Syllabus (subject to change)
Lecture 1, Mar 08 1999: Course overview, introduction to Machine
Learning, Concept Learning. Class-notes available as:
- postscript, chapter 1,
- pdf (adobe acrobat), chapter 1,
- postscript, chapter 1 (4 slides to a page),
- pdf (adobe acrobat), chapter 1 (4 slides to a page),
- postscript, chapter 2,
- pdf (adobe acrobat), chapter 2.
Lecture 2, March 15 1999:
More on concept learning, Decision Trees
Class-notes available as:
Reading (distributed in class):
- Frontiers of Finance Economist, October 9 1993; Ridley, M.
(survey article on finding patterns in financial markets)
- Inducing Stock Screening Rules for Portfolio Construction Journal of the Operational Research Society Vol. 42 No. 9 1991; Tam, K. Y. Kiang, M. Y. and Chi, R. T. H. (decision trees for stock screening)
Mar 22 1999 NO CLASS TODAY
Lecture 3, Mar 29 1999
Decision Trees: complete class notes, discuss paper Inducing Stock Screening Rules for Portfolio Construction, beginning of neural networks
Classnotes available as:
Lecture 4, Apr 5 1999
Neural Networks continued
-- London will have this class on April 9th
Readings (distributed April 5th, may arrive April 7th):
- Constructive Learning and Its Application to Foreign Exchange Rate Forecasting A. N. Refenes (from "Neural Networks in Finance and Investing" Trippi and Turban Eds)
- Book chapter 14 from "C++ Neural Networks and Fuzzy Logic" Application to Financial Forecasting
Lecture 5, Apr 12, 1999 Guest Lecturer James Thomas -- Genetic Algorithms for Computational Finance
Readings (distributed April 5th, may arrive April 7th):
- Chapter 9 from text-book (please read before class)
- Is Technical Analysis in the Foreign Exchange Market Profitable? A Genetic Programming Approach Neely, Wellar and Ditmar (please read before class)
- Technical Trading Rule Profitability and Foreign Exchange Intervention, Baron (optional back-ground reading).
2 sets of class-notes available as:
Lecture 6, Apr 19, 1999 Instance Based Learning (locally weighted regression, k-nearest neighbor)
--This class will be taught live in New York on Apr 19, live in London on Wednesday Apr 21
Class-notes available as:
We will also start Hidden-markov models and EM in this class.
(Some notes on HMM and EM, which may or may not be relevant, and which are not the ones presented in class, can be found at here)
Lecture 7, Apr 26, 1999
- Exam (in class)
Final exam in ML class will be this coming Monday, April 26, at the
usual class time. Open books, open notes. Please bring your own
calculator.
rosie+46838 @ cs.cmu.edu
Last modified: Wed May 12 13:59:30 EDT 1999