Homework

1
Homework

• Homework 1 is up
• Programming Language whatever
• Write your own code
• HW questions about code Be succinct and clear
• Partial credit will be given
• So briefly justifying your answers may help

2
Look at the homework early, because

• No class next Monday
• Your TA is out next week
• In the meantime, were both available via e-mail

3
Project Guidelines

• Project proposal due October 16 (1 pg)
• Who is in your group
• Your task (and why is it interesting?)
• Where did/will you get your data?
• Which ML algorithms will you try first?
• Final project write-up due December 8th
• Web page
• Report (4 pgs, ACM formatlink on course page)

4
Some project ideas

• The standard problems
• Handwriting, text classification, disease
  detection, etc. see the UCI ML repository
• Recommendations
• E.g., the Netflix prize
• Sports predictions
• Question how does intransitivity impact ML?
• Multi-task learning
• TinyGrams
• Google n-grams corpus gives P(phrase) for up to
  5-word phrases
• Based on 1 trillion words unprecedented coverage
• But around 150G uncompressed could an ML
  approximation fit in memory?

5
More researchy projects

• A couple of project ideas related to information
  extraction
• Example TextRunner
• If interested, drop me an e-mail (soon)