Description
Question 1: Game Tree Search [60 points]
Consider the game tree below. Let ∆ and ∇ nodes represent nodes belonging to the maximizing and
minimizing player respectively.
A
B
D
3 4
E
7 2
C
F
8 1
G
0 6
1. (20 points) Suppose we wish to run the Minimax algorithm on this game tree. Provide the game
theoretic values of nodes A . . . G through optimal play.
2. (10 points) Use alpha-beta pruning to compute the Minimax value at each node for the original game
tree (i.e., deterministic environment). Assume that nodes are visited from left to right. In addition to
the Minimax values, show your alpha and beta values at each node after the final time it is visited.
3. (5 points) Which branches, if any, in the game tree are pruned? Show this on the graph above.
4. (5 points) Explain, in English, why we would want to use alpha-beta pruning.
5. (5 points) Now suppose we are in a non-deterministic environment with rules as follows. When a player
begins their turn they first flip a fair coin (heads probability .5) and if it turns up heads they choose
the optimal action. However, if the coin turns up tails they then flip another fair coin to randomly
select their action. In this game, what is the probability of a player choosing an optimal action at each
node?
6. (5 points) Given the game described in part 5, redraw the game tree as follows. For each non-terminal
node X, introduce a new chance node X0
. Let X0 be the only successor of X, while the original
successors of X are now given to X0
. Label the probability on each edge from X0
to its successors.
These probabilities should sum to 1 for each X0
.
7. (10 points) Using the new game tree in part 6, solve the non-deterministic game. Specifically, provide
the expected game theoretic value of nodes A . . . G for the new tree. Round your answers to the nearest
2 decimal places (i.e. X.XX). Hint: this is explained on slide p.58.
Question 2: Natural Language Processing [40 points, 5 points each]
This question will require you to do some minimum coding or scripting to get the answers, but it is not a
programming question. Do not turn in any code. Instead, simply put your answers in the same pdf file as
the other questions. You may use any programming language or tools.
Download the zip file http://pages.cs.wisc.edu/~jerryzhu/cs540/handouts/WARC201709.zip. Unzip it on your computer, you should see 363 text files. These are your essays from homework 1. We call the
whole collection the corpus.
1. Briefly describe in English, the most convenient way you can think of for a computer program to break
the corpus into “computer words.” For example, if you use Java you may consider Scanner; if you use
unix commandline you may use wc. We want you to use the simplest method that you can think of.
It is OK if these computer words do not fully agree with our natural language definition of words: For
example, you may have a computer word like However, with that comma attached. Your method is a
crude “natural language tokenizer.” When we mention “word” we mean your computer word.
2. Under your method, how many computer word tokens (occurrences) are there in the corpus? How
many computer word types (distinct words) are there in the corpus?
3. Sort the word types by their counts in the corpus, from large to small. List the top 20 word types and
their counts.
4. Pick 20 bottom word types (they should all have count 1) and list them. You may want to include
some strange ones, if any.
5. Let the word type with the largest count be rank 1, the word type with the second largest count be
rank 2, and so on. If multiple word types have the same count, you may break the rank tie arbitrarily.
This produces (r1, c1), . . . ,(rn, cn) where ri
is the rank of the ith word type, and ci
is the corresponding
count of that word type in the corpus; n is the number of word types. Plot r on the x-axis and c on
the y-axis, namely each (ri
, ci) is a point in that 2D space (you can choose to connect the points or
not).
6. Plot log(r) on the x-axis and log(c) on the y-axis. You may choose the base.
7. Briefly explain what the shape of the two curves mean.
8. Discuss two potential major issues with your computer words, if one wants to use them for natural
language processing.
