80-629 -- Homework

Due Date: Friday October 12, 2018
Handing in: by email to the instructor (laurent.charlin _at_ hec.ca). Subject: [80-629] HW report

This homework will be worth 20% of your final grade. You are more than welcome to exchange and discuss ideas related to your work with colleagues in the class but the final report must be your own.

Grading Scheme:

The problem is clearly introduced	(5)
The report is technically correct	(5)
Literature is sufficiently surveyed / Model is correctly compared to or contrasted with others / Application is innovative	(5)
The report has sufficient technical depth	(5)
Total	(20)

Todo: Pick one of the three options below and write (a maximum of) four page technical report on it (you are allowed unlimited references in addition to the four pages). Your target audience are the students in the class, that is other students in the class should be able to read and understand your report.

1. Literature review: Summarize the relevant literature pertaining to a machine learning topic. The topic can be a fundamental research area in machine learning or an application for machine learning (e.g., bioinformatics or finance). The topic can be as narrow or as wide as you want. In the former case I would expect that you would review recent research papers on the topic. In the latter case you could take one (or two) main sources (e.g., chapters of a book on the topic).
2. Model: Explain a machine learning model that we have not discussed in class. Be precise about it mathematically, also make sure that you provide intuition about what it is trying to accomplish. Compare and contrast to other models (e.g., advantages/disadvantages, where can it be used, complexity of fitting procedure). Empirical experiments can be helpful to discuss specific properties of the model but are not expected.
3. New application: Pick an application that has not received attention from the machine learning field. Describe the application and describe how machine learning could help. Try to be specific about the expected benefit of machine learning. Make sure that you are specific: clearly define the task, the performance measure, and the experience, describe how machine learning models could learn from the data and clearly specify inputs and outputs.

There is no standard report format but I would expect that regardless of which of the three option you select, you start your report with a brief introduction and follow up with an in-depth technical section. In the title indicate which option you chose (from the three above). Please make sure to correctly cite work that is not your own.

It is perfectly fine to use this homework as the starting point for your class project. Note however that you cannot use the same material twice.