Description
Question 1 (30 points)
For this question you will write a program to solve TSP using simulated annealing. The setup
for the problem is exactly the same as for the A∗
-search question in Assignment 1, and you are
allowed to re-use any code you would like from your Assignment 1 submission. The data set for
this problem is also identical to that used in Assignment 1, and I have posted a copy in the file
TSPproblem1.zip in the Assignment 2 folder.
1
CS 486/686: Introduction to Artificial Intelligence Assignment 2
a. Explain what local search operators you use (make sure that the operators preserve a tour). That
is, how do you generate the MoveSet or Neighbours? (2 pts)
b. Experiment with at least 3 different annealing or temperature schedules, and report your findings. What schedule would you chose to use? (10 pts)
c. Run the program on the 36-city TSP problem instance given in the file problem36. Plot how
the cost of the solution changes during one run of the algorithm. What is the cost of the best
solution that your algorithm found (do not let it run for more than 5 minutes)? (12 pts)
d. Is your version of simulated annealing complete? Explain why or why not. (2 pts)
e. Is it optimal? Explain why or why not. (2 pts)
Then we give 2 pts for well documented code.
Submit the following;
• A well documented copy of your code.
• A discussion of the local search operators you used.
• A discussion of the annealing schedules you used and how they influenced your search results.
• Your well-labeled plot from question c.
• Answers to questions posed in d and e.
Question 2 (45 points)
This question looks long but Don’t Panic!. Most of this question is detailed documentation!
In this question, you will design an agent that plays the popular game of “Connect Four”1
. We
have provided an interactive framework for playing the game, including the rules of the game and
a rudimentary basic player utilizing minimax search. You will implement a better agent that uses
alpha-beta pruning and better evaluation functions.
If you are interested, as an optional component of this assignment, you may enter your agent
in a tournament against other submitted agents.
1This assignment and the source code for this Connect Four implementation are based on material from MIT
OpenCourseWare (OCW). They have been modified from the original distribution. This is distributed under a Creative
Commons License.
2
CS 486/686: Introduction to Artificial Intelligence Assignment 2
Notes
• We are providing code for this question. Download a2Q2.zip. You should find the following
basicplayer.py, connectfour.py, implementation,py, LISCENCE, main.py,
tests.py, tree searcher.py, util.py. The only file you need to modify is
implementation.py.
• You should use Python 3 for this assignment.
• Your answers for the programming portion belong in the file implementation.py. You
may change the other files for debugging purposes, but before you submit your assignment,
you should test with the other files in their original state. When your code is tested on our
end all files besides implementation.py will be replaced by the original copy.
• A number of basic tests have been provided for you in tests.py, to test basic competence
of your agent. Please note that they are by no means exhaustive, and passing them is not
a guarantee of good performance. Before running the full set of tests, run python -m
unittest tests.TestAlphaBetaSearch. Only after you’ve passed these tests
then should you go on and run the full tester, which may be accessed by running python
-m unittest.
• Remember: Your implementation.py must contain your student ID and a name for
your agent. To do this, uncomment and edit the appropriate lines at the start of the file. If
you wish to participate in an optional tournament with other agents, please set COMPETE to
True.
• In your writeup file (e.g. writeup.pdf), you will include a short description of the “better”
evaluation function you create and document your design choices for it (see Part (c)).
• Submit: Upload to the LEARN Dropbox your report and your implementation.py
file. If you have additional .py files that are required for your code to run, you may submit
those as well.
Introduction to Connect Four
This question is about adversarial search, and it will focus on the game “Connect Four”.
A board is a 7×6 grid of possible positions. The board is arranged vertically: 7 columns, 6 cells
per column, as follows:
0 1 2 3 4 5 6
0 * * * * * * *
1 * * * * * * *
3
CS 486/686: Introduction to Artificial Intelligence Assignment 2
2 * * * * * * *
3 * * * * * * *
4 * * * * * * *
5 * * * * * * *
Two players take turns alternately adding tokens to the board. Tokens can be added to any column that is not full (ie., does not already contain 6 tokens). When a token is added, it immediately
falls to the lowest unoccupied cell in the column.
The game is won by the first player to have four tokens lined up in a row, either vertically:
0 1 2 3 4 5 6
0
1
2 O
3 O X
4 O X
5 O X
horizontally:
0 1 2 3 4 5 6
0
1
2
3
4
5 X X X O O O O
or along a diagonal:
0 1 2 3 4 5 6
0
1
2 O
3 O O X
4 O O X X
5 O O X X X
4
CS 486/686: Introduction to Artificial Intelligence Assignment 2
You can get a feel for how the game works by playing it against the computer. For example,
you can play O, while a computer player that does minimax search to depth 4 plays X: python
main.py O. See python main.py -h for a list of all options.
For each move, the program will prompt you to make a choice, by choosing what column to
add a token to.
The prompt may look like this:
Player 1 (X) puts a token in column 2
0 1 2 3 4 5 6
0
1
2
3
4
5 X
Pick a column #: –>
In this game, Player 1 just added a token to Column 2. The game is prompting you, as Player
2, for the number of the column that you want to add a token to. Say that you wanted to add a
token to Column 1. You would then type ‘1’ and press Enter.
The computer, meanwhile, is making the best move it can while looking ahead to depth 4 (two
moves for itself and two for you).
The Code
ConnectFourBoard
connectfour.py contains a class entitled ConnectFourBoard. ConnectFourBoard
objects are immutable. To make a move on a board, you create a new ConnectFourBoard object with your new token in its correct position. This makes it much easier to make a search tree of
5
CS 486/686: Introduction to Artificial Intelligence Assignment 2
boards: You can take your initial board and try several different moves from it without modifying
it, before deciding what you actually want to do. The provided minimax search takes advantage
of this; see the get all next moves, minimax, and minimax find board value functions in basicplayer.py.
So, to make a move on a board, you could do the following:
>>> myBoard = ConnectFourBoard()
>>> myBoard
0 1 2 3 4 5 6
0
1
2
3
4
5
>>> myNextBoard = myBoard.do_move(1)
>>> myNextBoard
0 1 2 3 4 5 6
0
1
2
3
4
5 X
You are welcome to call any methods associated with the object ConnectFourBoard. However, these are the key ones that you will use the most:
• ConnectFourBoard() (the constructor) – Creates a new ConnectFourBoard instance.
You can call it without any arguments, and it will create a new blank board for you.
• get current player id() – Returns the player ID number of the player whose turn it
currently is.
6
CS 486/686: Introduction to Artificial Intelligence Assignment 2
• get other player id() – Returns the player ID number of the player whose turn it
currently isn’t.
• get cell(row, col) – Returns the player ID number of the player who has a token in
the specified cell, or 0 if the cell is currently empty.
• get top elt in column(column) – Gets the player ID of the player whose token is
the topmost token in the specified column. Returns 0 if the column is empty.
• get height of column(column) – Returns the row number for the highest-numbered
unoccupied row in the specified column. Returns -1 if the column is full, returns 6 if the
column is empty. NOTE: this is the row index number not the actual ”height” of the column,
and that row indices count from 0 at the top-most row down to 5 at the bottom-most row.
• do move(column) – Returns a new board with the current player’s token added to column. The new board will indicate that it’s now the other player’s turn.
• longest chain(playerid) – Returns the length of the longest contiguous chain of
tokens held by the player with the specified player ID. A ’chain’ is as defined by the Connect
Four rules, meaning that the first player to build a chain of length 4 wins the game.
• chain cells(playerid) – Returns a Python set containing tuples for each distinct
chain (of length 1 or greater) of tokens controlled by the current player on the board.
• num tokens on board() – Returns the total number of tokens on the board (for either
player). This can be used as a game progress meter of sorts, since the number increases by
exactly one each turn.
• is win() – Returns the player ID number of the player who has won, or 0.
• is game over() – Returns true if the game has come to a conclusion. Use is win to
determine the winner.
Other Useful Functions
• get all next moves(board) in basicplayer.py – Returns a generator of all moves
that could be made on the current board
• is terminal(depth, board) in basicplayer.py – Returns true if either a depth
of 0 is reached or the board is in the game over state.
• run search function(board, search fn, eval fn, timeout) in util.py
– Runs the specified search function with iterative deepening, for the specified amount of
time. Described in more detail below.
7
CS 486/686: Introduction to Artificial Intelligence Assignment 2
• human player in connectfour.py – A special player, that prompts the user for where
to place its token. See below for documentation on ”players”.
• run game(player1, player2, board = ConnectFourBoard()) in connectfour.py
– Runs a game of Connect Four using the two specified players. ’board’ can be specified if
you want to start off on a board with some initial state.
To play a game, you need to turn a search algorithm into a player. A player is a function that
takes a board as its sole argument, and returns a number, the column that you want to add a piece
to. You can define a basic player as follows:
def my_player(board):
return minimax(board, depth=3, eval_fn=focused_evaluate, timeout=5)
or, more succinctly (but equivalently):
my_player = lambda board: minimax(board, depth=3, eval_fn=focused_evaluate)
However, this assumes you want to evaluate only to a specific depth. In class, we discussed
the concept of iterative deepening. We have provided the run search function helper function to
create a player that does iterative deepening, based on a generic search function. You can create an
iterative-deepening player as follows:
my_player = lambda board: run_search_function(board,
search_fn=minimax, eval_fn=focused_evaluate)
Writing a ConnectFour Agent
The goal of this question is to create a Connect Four playing agent using alpha-beta pruning and
evaluation functions. We have provided you with code for an agent that uses minimax search with
a rather silly evaluation function. Can you beat it?
The first thing you need to do is create a better evaluation function than the one we provide.
An evaluation function takes one argument, an instance of ConnectFourBoard. They return
an integer that indicates how favourable the board is to the current player. The problem with the
current evaluation function (basic evaluate) is that it treats all winning positions the same,
no matter when they might occur (similarly with losing position) and this can sometimes lead to
odd behaviour of the program, and certainly does not reflect how people may play who tend to like
to reach winning positions sooner and losing positions later.
a. (Focused Evaluation Function) (5 points): In implementation.py, write a new evaluation function, focused evaluate, which prefers winning positions when they happen sooner
and losing positions when they happen later.
Here are some suggestions.
8
CS 486/686: Introduction to Artificial Intelligence Assignment 2
1. Look carefully at the basic evaluate function.
2. Leave the “normal” values (the ones that are like 1 or -2, not 1000 or -1000) alone. You
don’t need to change how the procedure evaluates positions that aren’t guaranteed wins or
losses.
3. Indicate a certain win with a value that is greater than or equal to 1000, and a certain loss
with a value that is less than or equal to -1000.
4. Remember, focused evaluate should be very simple and fast. The point of this exercise
is mostly to familiarize you with the code and the basic evaluation function structure. Later
in the assignment you are asked to develop something even more sophisticated.
You do not need to include your description of focused evalue in your final report.
b. Alpha-Beta Pruning (25 points): In implementation.py, write a procedure alpha beta search
which does alpha-beta pruning to reduce the search space used by minimax search.
Feel free to follow the model set by minimax, in basicplayer.py.
Your alpha beta search function must take the following arguments:
• board – The ConnectFourBoard instance representing the current state of the game
• depth – The maximum depth of the search tree to scan.
• eval fn – The ”evaluate” function to use to evaluate board positions
And optionally it takes two more function arguments:
• get next moves fn – a function that, given a board/state, returns the successor board/states.
By default, get next moves fn takes on the value of basicplayer.get all next moves.
• is terminal fn – a function that given a depth and board/state, returns True or False.
True if the board/state is a terminal node, and that static evaluation should take place. By
default is terminal fn takes on the value of basicplayer.is terminal.
You should use these functions in your implementation to find next board/states and check termination conditions.
The search should return the column number that you want to add a token to. If you are experiencing massive tester errors, make sure that you’re returning the column number and not the entire
board!
Other notes:
9
CS 486/686: Introduction to Artificial Intelligence Assignment 2
• Write a procedure alpha beta search, which works like minimax, except it does alphabeta pruning to reduce the search space.
• You should always return a result for a particular level as soon as alpha is greater than or
equal to beta.
• util.py defines two values INFINITY and NEG INFINITY, equal to positive and negative infinity; you should use these for the initial values of alpha and beta.
• This procedure is called by the player alphabeta player. This is defined in implementation.py.
• Your procedure will be tested with games and trees, so don’t be surprised if the input doesn’t
always look like a Connect 4 board.
c. A Better Evaluation Function (15 points):
Your goal is to implement a new procedure for static evaluation that outperforms the evaluation
function in the basic player provided to you; this procedure is called better evaluate
and a stub is provided for you in implementation.py. It is evaluated by the test case TestConnectFourPlay, which plays my player against basic player in a tournament of 4 games.
Clearly, if you just play basic player against itself, each player will win about as often as it
loses. We want you to do better with a goal of producing a player that wins at least twice as often
as it loses.
Once you have designed the ultimate Connect Four evaluation function, include a short (max
500 word) description of your design choices. In particular, you should explain how your evaluation function works and why you believe it to be the best Connect Four agent.
A couple of additional notes:
• Your agent should win legitimately! It can not interfere with the other player.
• Remember to cite your references! There is a lot of information about the game of Connect
Four available. You are welcome to read any of the work you like, but you must cite it.
Citatations can be either in the written report or included in your code documentation. You
may not use code that accesses online resources, not can you directly copy or paste someone
else’s implementation of alpha-beta search.
Advice
The rule of thumb with evaluation functions is that simpler is often better. Note that your time
is limited. Given the choice between using that time to search deeper vs using it on a complex
evaluation function, searching deeper is almost always better.
10
CS 486/686: Introduction to Artificial Intelligence Assignment 2
The full set of tests can take a long time to run. You can pick to only run a subset of tests, for
example, python -m unittest tests.TestAlphaBetaSearch. You may then run the
full tests via python -m unittest.
Important Notes
• After you’ve implemented better evaluate, please comment out the following line in
your implementation.py:
better evaluate = memoize(basic evaluate)
and then uncomment the following line:
better evaluate = memoize(better evaluate)
The original setting was set to allow you play the game initially, but it is unnecessary and
incorrect once you’ve implemented your version of better evaluate.
• Ensure you have entered your name and a name for your agent on the appropriate lines at
the top of implementation.py.
• Set COMPETE to True (in implementation.py) if you would like to participate in a
tournament. If there is enough interest then we will run a small tournament. This is just for
fun and is entirely optional. No marks are awarded for participating. Similarly there are no
penalties if you decide not to enter of if you enter and your agent does poorly.
Question 3 (25 points)
Equine colic is one of the leading causes of death in adult horses. However, if diagnosed early
enough, it is often surgically curable. Using the language of your choice, write a decision tree
algorithm that will learn to diagnose whether a patient is healthy or has colic. Use the horseTrain
file, (found in the Assignment 2 folder) to train the decision tree. Each training instance has 16 numeric attributes (features) and a classification, all separated by commas. The attributes correspond
to the following measurements made from each patient at admission to the clinic.
1. K
2. Na
3. CL
4. HCO3
5. Endotoxin
6. Aniongap
11
CS 486/686: Introduction to Artificial Intelligence Assignment 2
7. PLA2
8. SDH
9. GLDH
10. TPP
11. Breath rate
12. PCV
13. Pulse rate
14. Fibrinogen
15. Dimer
16. FibPerDim
In the decision tree, use only binary tests, i.e. each node should test whether a particular
attribute has a value greater or smaller than a threshold. In deciding which attribute to test at
any point, use the information gain metric. Set the node test threshold for each potential attribute
using this same metric i.e. at each point, see all the values that exist for a particular attribute in
the remaining instances, order those values, and try threshold values that are (half way) between
successive attribute values. Use the threshold value that gives the highest information gain. Allow
the same attribute to be tested again later in the tree (with a different threshold). This means that
along a path from the root to a leaf, the same attribute might be tested multiple times. After learning
the decision tree, use the horseTest file to test the generalization accuracy of the tree.
Submit the following
1. (5 points) Your code and documentation.
2. (15 points) Picture of the decision tree. Hand drawn and then scanned is fine.
3. (2 points) Answer the question “How many of the training instances does the tree classify
correctly?”
4. (3 points) Answer to the question “How many of the test instances does the tree classify
correctly?”
12
CS 486/686: Introduction to Artificial Intelligence Assignment 2
Format of data files
We are providing you with data from an actual vet clinic. Each line in a file is one instance. The
first 16 numbers are the values for the 16 attributes listed above. The last entry on a line is whether
the horse was healthy or not. You are allowed to reformat the data files in what ever way you want
so as to make reading them easier for your program.
13
