CS221 Problem Workout

Week 6

1) Problem 1: Dinner Round Table CSP

(a) You and your friends (Veronica, Jarvis, Gabriela, Kanti) sit at a round table for
dinner as follows, with you (Y ) at the top, followed by Jarvis (J), Kanti (K),
Gabriela (G), and Veronica (V ) in clockwise order: There are three dishes on

the menu: the vegetarian deep dish pizza, the chicken quesadilla, and the beef
cheeseburger, and each person plans to order exactly one dish.
Unfortunately, rather than simply enjoying the dinner, the event turns into a
logistical nightmare as you and your friends impose all the following constraints
upon yourselves:
i. You (Y ) are vegetarian.
ii. If Veronica (V ) orders beef, then Jarvis (J) will order veggie, and vice versa.
iii. Kanti (K) and Jarvis (J) do not want to both get non-chicken dishes.
iv. Each person wants to order something different than what the two friends
sitting next to them order. For instance, Veronica (V ) wants to order a
different meal than you (Y ) and Gabriela (G).
Formulating this as a constraint satisfaction problem:
• What are the variables X = (X1 , ..., Xn ) based on the round table diagram
above? What are the values that can be assigned to each variable?
• What are the factors between the variables? Draw edges into the round
table diagram above for each factor between variables, and label each edge

1
 with an expression that represents the constraint between the variables (e.g.,
[Y = Veggie]).
• What values are removed from the domain once arc consistency is enforced
in both directions for each pair of variables? Cross out the removed values in
the round table diagram above to enforce arc consistency.

(b) Your server comes by your table and says that they are out of beef today, so you
and your friends decide to rework your constraints. Now they are:
i. You (Y ) are vegetarian.
ii. Each person wants to order something different than what the two friends
sitting next to them order.
Is this new constraint problem satisfiable? For convenience, the updated table is
given below.

2
(c) Realizing that the constraints are too strict, your group decides to relax them to
the following:
i. You (Y ) are vegetarian.
ii. Instead of hard requiring every adjacent person to order something different,
it is now only a preference. To represent this, you define factors:
• f1 (Y, J) = 1[Y ̸= J] + 1
• f2 (J, K) = 1[J ̸= K] + 1
• f3 (K, G) = 1[K ̸= G] + 1
• f4 (G, V ) = 1[G ̸= V ] + 1
• f5 (V, Y ) = 1[V ̸= Y ] + 1
How does defining the factors this way make the problem satisfiable? What
is the maximum weight?

3
2) Problem 2: N-Queens CSP
The N-Queens problem is a classic puzzle involving the placement of N chess queens
on an N × N chessboard so that no two queens threaten each other. You can find
a detailed description of the 8-Queens variant and then a generalization to N-Queens
here on Wikipedia if you’re not too familiar with chess. As it turns out, the N-Queens
problem leads itself very well into the formulation of a constraint satisfaction problem.
We’ll approach the problem from the angle of analyzing the chess board one row at
a time, as no two queens can be on the same row (or they threaten each other). Let
our variables be the position of each queen piece Q = Q1 , ..., QN on each row, e.g. the
queen piece on row 1 has the variable Q1 . Let the possible values for a position be the
column numbers 1, ..., N , e.g. if we were to place the queen piece of row 1 in column
N, then we would do the assignment: Q1 = N .

(a) Using the variable definitions we defined above, express the constraints in this
problem as factors, such that an assignment is consistent only if the following
constraints of the problem are satisfied:
• No two queens are in the same column
• No two queens are on the same diagonal of the chess board

(b) Suppose N = 4, and we want to find a solution to the 4-Queens problem through
backtracking search. We’ll use the convention of starting our assignments from
the top-most row 1, and go left to right from column 1 to 4, before moving on to
the next row, with the last row 4 at the bottom of the board. What solution do
we get for 4-queens if we were to use backtracking search with this convention?
For now, just use plain backtracking; don’t worry about forward checking.

4
(c) During your backtracking search, you should’ve noticed that assigning Q1 = 1
leads to no possible assignments for the other queens that would satisfy the prob-
lem. Repeat the backtracking search for the Q1 = 1 branch, but with forward
checking this time. Trace out the domain of values for our variables Q2 , Q3 , Q4 at
each step. You should notice that Q1 = 1 is not a viable assignment sooner with
forward checking than without.
Food for thought: if you had to code up backtracking search with forward check-
ing, then what is one complication you’d have to account for with the pruned
domain when going back up a branch?

5
 3) (optional) Problem 3: Farm Setup CSP
Farmer Kim wants to install a set of sprinklers to water all his crops in the most cost-
effective manner and has hired you as a consultant. Specifically, he has a rectangular
plot of land, which is broken into W × H cells. For each cell (i, j), let Ci,j ∈ {0, 1}
denote whether there are crops in that cell that need watering. In each cell (i, j), he can
either install (Xi,j = 1) or not install (Xi,j = 0) a sprinkler. Each sprinkler has a range
of R, which means that any cell within Manhattan distance of R gets watered. The
maintenance cost of the sprinklers is the sum of the Manhattan distances from each
sprinkler to his home located at (1, 1). Recall that the Manhattan distance between
(a1 , b1 ) and (a2 , b2 ) is |a1 −a2 |+|b1 −b2 |. Naturally, Farmer Kim wants the maintenance
cost to be as small as possible given that all crops are watered. See figure below for
an example.

Figure 1: An example of a farm with W = 5 and H = 3. Each cell (i, j) is marked ‘C’ if
there are crops there that need watering (Ci,j = 1). An example of a sprinkler installation
is given: a cell (i, j) is marked with ‘S’ if we are placing a sprinkler there (Xi,j = 1). Here,
the sprinkler range is R = 1, and the cells that are shaded are the ones covered by some
sprinkler. In this case, the sprinkler installation is valid (all crops are watered), and the total
maintenance cost is 1 + 4 = 5.

Farmer Kim actually took CS221 years ago, and remembered a few things. He says: “I
think this should be formulated as a factor graph. The variables should be Xi,j ∈ {0, 1}
for each cell (i, j). But here’s where my memory gets foggy. What should the factors
be?” Let X = {Xi,j } denote a full assignment to all variables Xi,j . Your job is to
define two types of factors:

• fi,j : ensures any crops in (i, j) are watered,

so that a maximum weight assignment corresponds to a valid sprinkler installation

6
fi,j (X) =

fcost (X) =