CS 151: Artificial Intelligence
Homework 5
Due date: 11/14/13 by 11:59 pm

In this assignment, you'll find, read and summarize/review one journal article or conference publication related to what you would like to do for your final project. The goals of this assignment are to give you more exposure to current AI research, have you practice your research skills, have you practice reviewing research papers, and (perhaps most importantly) help you find some background information for your final project.

It is strongly recommended that you work in pairs or triples for the final project!

1. Finding your literature

   Your first task is to select the paper you will read and review for this assignment related to your final project idea. If you are working in a team for your final project, each of you should read a different article (closely related but different) so that all of you get some different background knowledge.

   If you're stuck on where to find papers or topics, look on the Resources page

2. Read and Summarize/Review Articles

   The next phase of this assignment is the review itself. Your task is to write a critical review of your paper. Each review should be 1-2 pages in length single-spaced. The format you should use is a short-answer rather than long-essay format. That is, you should specifically address each of the points below. In your writeup you should list the point, and then give a concise, informative response to the point. You should write clean, coherent, well-structured text. Just because you are answering specific questions does not mean you can forget about grammar, organization, etc.

   Here are the points you must address in your review:

   • Summarize the work in 1 paragraph
   • (State concisely) What problem are the authors trying to solve?
   • What are this work's main contributions?
   • Is this work technically sound? (i.e., How good is the solution?  How convinced are you that it works?)
   • Comment on the quality of the writing.
   • What (and how much) impact will this work have (has it had) on its field? (Note: Because these papers have already been published, they hopefully have already had an impact on the their field. If your research in looking for a paper revealed the impact of this paper's work, state it. Otherwise, try to "predict" what that impact was/might be - backing up your prediction with reasoning.)

   I will be grading your review with a letter grade - just as you would receive on a paper. To help you write this review, here are some general and specific tips:

   • Back up your claims by citing specific examples from the paper whenever possible (not necessarily quotes, but citing paragraph 2 of the related work section for example).
   • Write a rough draft, then go back and revise it just as you would any essay or paper.
   • Imagine that your comments will be read by the authors of the paper. Make your comments as constructive/helpful to them as possible.

Congratulations! You've made it to the last programming project of the semester. In this assignment, you will be asked to design a naive Bayes classifier, a perceptron, and a large-margin classifier (similar to a support vector machine). You will be using these classifiers to classify handwritten digits and human faces.

• You may work in pairs for this assignment as well. If you choose to work with someone else, submit only one directory. Put both of your names at the top of the bustersAgents.py and inference.py files. You can also put both of your written assignment pdfs inside this directory.


• The last part of this assignment asks you to design your own features: the individual or pair whose features give the best classification accuracy will be given extra credit.


• So that we're all working with the same version of the code, download the zipped directory here. If you're ready, click here


• In case the Berkeley website goes down, a copy of the assignment can be found here.

1. AIMA 18.6
2. AIMA 18.5
3. Let S be a set of training examples with p positive and n negative examples. Let A be some attribute. Is it always true that H(S) >= R(A) where H(S) is the entropy of S and R(A) is the expected entropy (i.e. the "remainder") after splitting on A?
4. AIMA 18.19
5. Construct a minimal-sized decision-tree that computes the XOR function of 2 inputs
6. Construct a support vector machine that computes the XOR function of two inputs (see AIMA 18.17)
7. Construct a neural networks by hand that computes the XOR function of two inputs (see AIMA 18.19)

You'll be turning in the "classification" directory which should contain your modified files. Add the pdf of your answers to the written questions (the book question and your literature review) to the "classification" directory. Give the pdf an intuitive name, e.g. "hw5_written_questions.pdf". Zip (compress) the "classification" directory and upload it using this URL.