Assignment 1

Due in class at 1:30pm, July 16. You may work in a group of up to three students, but each individual must be involved in every question. Do not assign problems to individuals within a team! Please submit only one solution per group.

We revisit the PGSS admissions committee data.

1. Try ID3 on the following data.

   	SAT	GPA	Essay	Performance
   Alice	good	good	good	good
   Bob	bad	good	good	bad
   Carrie	good	bad	bad	bad
   Eve	good	bad	bad	bad
   Spot	good	good	bad	good
   Krunch	bad	bad	good	???

   Clearly indicate all the gain computations you must do to build the decision tree, the final decision-tree hypothesis generated, and the resulting verdict for Krunch.

2. Try a perceptron on following data.

   	SAT	GPA	Essay	Performance
   Alice	1.0	1.0	0.5	1.0
   Bob	-1.0	1.0	0.5	-1.0
   Carrie	1.0	0.0	0.0	-1.0
   Spot	-1.0	-0.5	-0.5	1.0
   Krunch	-0.5	0.0	1.0	???

   The initial weights should be as follows.

   0.0	for the constant-1 input
   0.1	for the SAT
   0.1	for the GPA
   -0.2	for the essay

   Train the perceptron (using 0.1 for the training rate r) on Alice, then Bob, then Carrie, then Spot, each exactly once in order. List the weighted sum, the predicted label, and the updated weights after each. Then see what the perceptron would predict for Krunch. You might list your answer in a table, like the following.

   	initial	Alice	Bob	Carrie	Spot	Krunch
   weighted sum	---
   predicted label	---
   actual label	---	1	-1	-1	1	---
   constant-1 weight	0.0					---
   SAT weight	0.1					---
   GPA weight	0.1					---
   essay weight	-0.2					---