ML I, Week 9: Programming Assignment
We have covered in the Machine Learning Class the main learning tasks density
estimation, classification, and regression. In this assignment, you are asked
to implement a neural net for regression and/or classification and to apply
it to data from a realistic problem (e.g., drug design, biotechnology,
or finance) of your choice. Select an appropriate model using (cross-)validation.
Document the following:
-
A description of your implementation of regression/classification,
and model selection. (What technique did you implement? Why
did you choose that technique?). Write about 10 sentences.
-
A description of your data (What kind of data did you choose? Why those?
How many data points and dimensions? Is this data set used by other people?) Write
about 5-10 sentences.
-
A discussion of your application of the ML techniques to the data (What
are the results? Did you expect these results? How do they compare with
results from the literature (if available)? Did you observe any difficulties
with the ML technique and/or the data? Which difficulties and how did you
solve them?) Write about 3-5 pages.
-
A printout of your code.
Hand in your report to Sibylle Mueller on Thursday, January 23, in class.
Also, send your code (in a single file) by email to Sibylle Mueller (muellers@inf.ethz.ch) with
subject 'ML programming assignment'. You can work on your own or with
a partner.
Important: With this exercise, we intend that you gain experience with
the topics discussed in class AND that you learn to document your results
carefully, i.e., such that someone else can repeat your experiments. So, please
include a bibliography, web links, a discussion about strategy parameter
choice and motivation of their choice. Imagine you would publish
your results in a scientific journal.
due Thursday, January 23, in class