Lab 3: Constraint Satisfaction

From 6.034 Wiki

(Difference between revisions)

Jump to: navigation, search

Current revision

This lab is due by Thursday, October 1 at 10:00pm.

1 A Working Vocabulary
2 Part 1: Warm-up
3 Part 2: Writing a depth-first search solver
- 3.1 Benchmarks
4 Part 3: Forward checking streamlines search by eliminating impossible assignments
- 4.1 Finding inconsistent values in neighbors
- 4.2 Depth-first constraint solver with forward checking
5 Part 4: Propagation!
- 5.1 Domain reduction
- 5.2 Propagation through reduced domains
6 Part 5A: Generic propagation
7 Part 5B: A generic constraint solver
8 Part 6: Defining your own constraints
9 API
- 9.1 Constraint Satisfaction Problems
- 9.2 Constraint objects
10 Appendix: Setting up a Constraint Satisfaction Problem
11 FAQ
12 Survey

Before working on the lab, you will need to get the code. You can...

Use Git on your computer: git clone username@athena.dialup.mit.edu:/mit/6.034/www/labs/fall2020/lab3

Use Git on Athena: git clone /mit/6.034/www/labs/fall2020/lab3

Download it as a ZIP file: http://web.mit.edu/6.034/www/labs/fall2020/lab3/lab3.zip

View the files individually: http://web.mit.edu/6.034/www/labs/fall2020/lab3/

All of your answers belong in the main file lab3.py. To submit your lab to the test server, you will need to download your key.py file and put it in either your lab3 directory or its parent directory. You can also view all of your lab submissions and grades here.

A Working Vocabulary

Before beginning, you may want to (re)familiarize yourself with the following terms:

variable: something that can receive an assignment value
value: something that can be assigned
domain: a set of values
constraint: a condition that limits domains of possible values

Part 1: Warm-up

In this lab, you'll write programs that solve constraint satisfaction problems (CSPs). A CSP consists of variables, assignments, and constraints, and is represented by a ConstraintSatisfactionProblem object as described in the API.

First, we'll get familiarity with CSPs by writing a few helper routines.

has_empty_domains(csp): returns True if the supplied problem has one or more empty domains. Otherwise, returns False.
check_all_constraints(csp): returns False if the problem's assigned values violate some constraint. Otherwise, returns True.

Each function takes in an argument csp, which is a ConstraintSatisfactionProblem instance.

Part 2: Writing a depth-first search solver

Now you can use your helper functions to write the constraint solver:

def solve_constraint_dfs(problem) :

This is just like depth-first search as implemented in the search lab, but this time the items in the agenda are partially-solved problems instead of paths. Additionally, for this problem, we will also want to track the number of extensions so we can compare the different strategies for constraint propagation. At the end, instead of returning just a solution, you will return a tuple (solution, num_extensions), where

solution is the solution to this problem as a dictionary mapping variables to assigned values (see API for details); or None if there is no solution to the problem.
num_extensions is the number of extensions performed during the search. Recall that as before, an extension occurs whenever a problem is removed from the agenda for processing.

Here is a rough outline of how to proceed:

Initialize your agenda and the extension count.
Until the agenda is empty, pop the first problem off the list and increment the extension count.
If any variable's domain is empty or if any constraints are violated, the problem is unsolvable with the current assignments.
If none of the constraints have been violated, check whether the problem has any unassigned variables. If not, you've found a complete solution!
However, if the problem has some unassigned variables:
1. Take the first unassigned variable off the list using csp.pop_next_unassigned_var().
2. For each value in the variable's domain, create a new problem with that value assigned to the variable, and add it to a list of new problems. Then, add the new problems to the appropriate end of the agenda.
Repeat steps 2 through 6 until a solution is found or the agenda is empty.

Benchmarks

So that we can compare the efficiency of different types of constraint-satisfaction algorithms, we'll compute how many extensions (agenda dequeues) each algorithm requires when solving a particular CSP. Our test problem will be the Pokemon problem from 2012 Quiz 2, pages 2-4.

You can solve the Pokemon problem by calling solve_constraint_dfs(pokemon_problem) directly. Note that the Pokemon problem is already defined for you in test_problems.py. To get a copy of it, use the method get_pokemon_problem() in lab4.py.

Please answer the following questions in your lab file:

Question 1: How many extensions does it take to solve the Pokemon problem with just DFS?

Put your answer (as an integer) in for ANSWER_1.

Part 3: Forward checking streamlines search by eliminating impossible assignments

One problem with the solve_constraint_dfs algorithm is that it explores all possible branches of the tree. We can use a trick called forward checking to avoid exploring branches that cannot possibly lead to a solution: each time we assign a value to a variable, we'll eliminate incompatible or inconsistent values from that variable's neighbors.

Finding inconsistent values in neighbors

First, we will write a helper function to eliminate inconsistent values from a variable's neighbors' domains:

Suppose V is a variable with neighbor W. If W's domain contains a value w which violates a constraint with every value in V's domain, then the assignment W=w can't be part of the solution we're constructing — we can safely eliminate w from W's domain.

(Note that unlike the check_all_constraints function above, eliminate_from_neighbors checks all combinations of values, and is not restricted to comparing only variables that have assigned values.)

This function should return an alphabetically sorted list of the neighbors whose domains were reduced (i.e. which had values eliminated from their domain), with each neighbor appearing at most once in the list. If no domains were reduced, return an empty list; if a domain is reduced to size 0, quit and immediately return None. This method should modify the input CSP. Hint: You can remove values from a variable's domain using csp.eliminate(var, val). But don't eliminate values from a variable while iterating over its domain, or Python will get confused!

def eliminate_from_neighbors(csp, var) :

We strongly suggest working out examples on paper to get a feel for how the forward checker should find inconsistent values.

To reduce the amount of nested for-loops and to make debugging easier, you may find it helpful to write a small helper function that, for example, takes in two variables V and W, and two values v and w in their respective domains, and checks if there are any constraint violations between V=v and W=w.

Depth-first constraint solver with forward checking

Now, we will write our improved CSP solver which uses eliminate_from_neighbors above to apply forward checking while searching for variable assignments.

def solve_constraint_forward_checking(problem) :

The implementation for this function will be very similar to that of solve_constraint_dfs, except now the solver must apply forward checking (eliminate_from_neighbors) after each assignment, to eliminate incompatible values from the assigned variable's neighbors.

Note that if eliminate_from_neighbors eliminates all values from a variable's domain, the problem will be recognized as unsolvable when it is next popped off the agenda: do not preemptively remove it from consideration.

Answer the following question in your lab4.py file:

Question 2: How many extensions does it take to solve the Pokemon problem with forward checking?

Put your answer (as an integer) in for ANSWER_2.

Part 4: Propagation!

Forward checking is a useful tool for checking ahead for inconsistencies and reducing the search space. However, in many situations, it's ideal to prune inconsistent states even faster.

Domain reduction

A far-reaching strategy called domain reduction eliminates incompatible values not just between neighbors, but across all variables in the problem. You can apply domain reduction either before search (this is what Sudoku players do when they narrow down options before tentatively guessing a value) or after assigning each variable during search (as a more powerful variation of forward-checking).

As it turns out, the implementation for both of these are effectively identical:

Establish a queue. If using domain reduction during search, this queue should initially contain only the variable that was just assigned. If before search (or if no queue is specified), the queue can contain all variables in the problem. (Hint: csp.get_all_variables() will make a copy of the variables list.)
Until the queue is empty, pop the first variable var off the queue.
Iterate over that var's neighbors: if some neighbor n has values that are incompatible with the constraints between var and n, remove the incompatible values from n's domain. If you reduce a neighbor's domain, add that neighbor to the queue (unless it's already in the queue).
If any variable has an empty domain, quit immediately and return None.
When the queue is empty, domain reduction has finished. Return a list of all variables that were dequeued, in the order they were removed from the queue. Variables may appear in this list multiple times.

Note that when the queue initially contains only the assigned variable, the first step of propagation is just forward checking of the assigned variable's neighbors. "Propagation" occurs as we add more variables to the queue, checking neighbors of neighbors, etc.

You will now implement domain_reduction, which applies forward checking (checking for neighboring values' inconsistencies) with propagation through any domains that are reduced.

Recall that domain reduction utilizes a queue to keep track of the variables whose neighbors should be explored for inconsistent domain values. If you are not explicitly provided a queue from the caller, your queue should start out with all of the problem's variables in it, in their default order.

When doing domain reduction, you should keep track of the order in which variables were dequeued; the function should return this ordered list of variables that were dequeued.

If at any point in the algorithm a domain becomes empty, immediately return None.

def domain_reduction(csp, queue=None) :

This method should modify the input CSP.

Hint: You can remove values from a variable's domain using csp.eliminate(var, val). But don't eliminate values from a variable while iterating over its domain, or Python will get confused!

Answer the following question in your lab4.py file:

Question 3: How many extensions does it take to solve the Pokemon problem with DFS (no forward checking) if you do domain reduction before solving it?

Put your answer (as an integer) in for ANSWER_3.

Propagation through reduced domains

Now we'll see how we can take advantage of domain reduction during the search procedure itself.

When used during search, domain reduction makes use of the assignments you've made to progressively reduce the search space. The result is a new, faster, CSP solution method: propagation through reduced domains. After each assignment, propagation through reduced domains uses the domain_reduction subroutine to "propagate" the consequences of the assignment: to neighbors, then to neighbors of neighbors, and so on.

def solve_constraint_propagate_reduced_domains(problem) :

Note that if domain_reduction eliminates all values from a variable's domain, the problem will be recognized as unsolvable when it is next popped off the agenda: do not preemptively remove it from consideration.

Debugging hint: be sure to look at the return types of functions that you call!

Answer the following question in your lab4.py file:

Question 4: How many extensions does it take to solve the Pokemon problem with forward checking and propagation through reduced domains? (Don't use domain reduction before solving it.)

Put your answer (as an integer) in for ANSWER_4.

Part 5A: Generic propagation

The domain_reduction procedure is comprehensive, but expensive: it eliminates as many values as possible, but it continually adds more variables to the queue. As a result, it is an effective algorithm to use before solving a constraint satisfaction problem, but is often too expensive to call repeatedly during search.

Instead of comprehensively reducing all the domains in a problem, as domain_reduction does, you can instead reduce only some of the domains. This idea underlies propagation through singleton domains — a reduction algorithm which does not detect as many dead ends, but which is significantly faster.

Instead of again patterning our propagation-through-singleton-domains algorithm off of domain_reduction, we'll write a fully general propagation algorithm called propagate that encapsulates all three checking strategies we've seen: forward checking, propagation through all reduced domains, and propagation through singleton domains.

The function propagate will be similar to the propagation algorithms you've already defined. The difference is that it will take an argument enqueue_condition_fn, a function that takes a problem and a variable, and outputs whether the variable should be added to the propagation queue.

def propagate(enqueue_condition_fn, csp, queue = None) :

Propagation through singletons is like propagation through reduced domains, except that variables must pass a test in order to be added to the queue:

In propagation through singleton domains, you only append a variable to the queue if it has exactly one value left in its domain.

Common misconception: Please note that propagation never assigns values to variables; it only eliminates values. There is a distinction between variables with one value in their domain, and assigned variables: a variable can have one value in its domain without any value being assigned yet.

As a review, propagation eliminates incompatible options from neighbors of variables in the queue. When used during search, the propagation queue initially contains only the just-assigned variable. The three enqueueing conditions we've seen are:

forward checking: never adds other variables to the queue
propagation through singleton domains: adds a neighboring variable to the queue if its domain has exactly one value in it
domain reduction / propagation through reduced domains: adds a neighboring variable to the queue if its domain has been reduced in size

Write functions that represent the enqueueing conditions (predicates) for each of these. Each predicate function below takes in a CSP and the variable in question, returning True if that variable should be added to the propagation queue, otherwise False.

 def condition_domain_reduction(csp, var) :

 def condition_singleton(csp, var) :

 def condition_forward_checking(csp, var) :

Part 5B: A generic constraint solver

Now, you can use propagate to write a generic constraint solver. Write an algorithm that can solve a problem using any enqueueing strategy. As a special case, if the enqueue_condition is None, default to ordinary dfs instead --- don't eliminate options from neighbors (don't use any forward checking or propagation) at all.

def solve_constraint_generic(problem, enqueue_condition=None) :

Answer the following question in your lab4.py file:

Question 5: How many extensions does it take to solve the Pokemon problem with forward checking and propagation through singleton domains? (Don't use domain reduction before solving it.)

Put your answer (as an integer) in for ANSWER_5.

Part 6: Defining your own constraints

In this section, you will create some constraint functions yourself.

Assuming m and n are integers, write a function that returns True if m and n are adjacent values (i.e. if they differ by exactly one) and False otherwise.

def constraint_adjacent(m, n) :

Also write one for being non-adjacent.

def constraint_not_adjacent(m, n) :

The following example shows how you build a constraint object that requires two variables — call them A and B — to be different.

example_constraint = Constraint("A","B", constraint_different)

Some constraint problems include a constraint that requires all of the variables to be different from one another. It can be tedious to list all of the pairwise constraints by hand, so we won't. Instead, write a function that takes a list of variables and returns a list containing, for each pair of variables, a constraint object requiring the variables to be different from each other. (You can model the constraints on the example above.) Note that for this particular constraint (the must-be-different constraint), order does NOT matter, because inequality is a symmetric relation. Hence, in you should only have one constraint between each pair of variables (e.g. have a constraint between A and B, OR have a constraint between B and A, but not both).

def all_different(variables) :

Note: You should only use constraint functions that have already been defined. Don't try to create a new constraint function and use it in this function, because our tester will get confused.

API

In this lab, we provide an API for representing and manipulating partial solutions to constraint satisfaction problems.

Constraint Satisfaction Problems

A ConstraintSatisfactionProblem is an object representing a partially solved constraint satisfaction problem. Its fields are:

variables: A list containing the names of all the variables in the problem, in alphabetical order.

domains: A dictionary associating each variable in the problem with its list of remaining values.

assignments: A dictionary. Each variable that has already been assigned a value is associated with that value here. When the problem is entirely solved, assignments contains the solution.

unassigned_vars: An ordered list of all the variables that still need to have a value assigned to them.

constraints: A list of the constraints between the variables in the problem. Each constraint is a Constraint object.

Note: While you may read any of the above variables, you should probably not modify them directly; instead, you should use the following API methods:

get_domain(var): Returns the list of values in the variable's domain.

set_domain(var, domain): Sets the domain of the variable to the specified list of values, sorted alphabetically/numerically.

set_all_domains(domains_dict): Sets the domains field to the specified dictionary. Does not sort domains.

get_all_variables(): Returns a list of all the variables in the problem.

get_all_constraints(): Returns a list of all the constraints in the problem.

pop_next_unassigned_var(): Returns first unassigned variable, or None if all variables are assigned. Modifies unassigned_vars list.

set_unassigned_vars_order(unassigned_vars_ordered): Given an ordered list of unassigned variables, sets unassigned_vars. (By default, the unassigned_vars list is initialized in alphabetical order.)

eliminate(var, val): Removes the value from the variable's domain, returning True if the value was found in the domain, otherwise False if the value wasn't found.

get_assignment(var): If the variable has been assigned a value, retrieve it. Returns None if the variable hasn't been assigned yet.

set_assignment(var, val): Sets the assigned value of the variable to val, returning a modified copy of the constraint satisfaction problem. Throws an error if val is not in the domain of var, or if var has already been assigned a value. For convenience, also modifies the variable's domain to contain only the assigned value.

constraints_between(var1, var2): Returns a list of all constraints between var1 and var2. Arguments that are None will match anything: for example, constraints_between('X',None) will return all constraints involving X and any other variable, and constraints_between(None, None) will return all of the constraints in the problem.; Note: For your convenience, the constraints returned will always be altered to match the order of the arguments you passed to this method. For example, csp.constraints_between(None, 'Y') will return a list of all constraints involving 'Y' — and the constraints will be altered so that 'Y' is their second variable (var2) in every case.

get_neighbors(var): Returns a list of all the variables that share constraints with the given variable, ordered alphabetically.

add_constraint(var1, var2, constraint_fn): Given two variables and a function to act as a constraint between them, creates a Constraint object and adds it to the constraints list. The function constraint_fn must be a binary predicate function: it takes two arguments (a value for the first variable, and a value for the second variable) and returns True if the values satisfy the constraint, or False otherwise.

add_constraints(list_of_constraints): Add a list of Constraint objects to the constraints list. Useful for when you want to add several constraints to the problem at once, rather than one at a time using .add_constraint.

copy(): Return a (deep) copy of this constraint satisfaction problem. This method is particularly useful because you will want to make a copy of the CSP every time you assign a value to a variable.

Constraint objects

A Constraint is a fairly basic object representing a constraint between two variables. A Constraint object has three fields:

var1: The first variable associated with this constraint

var2: The second variable associated with this constraint

constraint_fn

A function that takes in two arguments, returning True or False depending on whether or not the given constraint is satisfied by the two arguments. For example,

constraint_equal(a, b) is a function requiring that a and b are equal.
constraint_different(a, b) is a function requiring that a and b are not equal.

These two functions are already defined in constraint_api.py, and can be accessed directly from lab4.py.

A Constraint object has just one method associated with it:

check(val1, val2): Applies this object's constraint_fn to two values (not variables), returning True if the values satisfy the constraint, or False otherwise.

Note: Due to certain limitations in our tester, a Constraint object constructor must take a named constraint_fn as an argument, NOT a lambda function.

Appendix: Setting up a Constraint Satisfaction Problem

The Pokemon problem from 2012 Quiz 2, pages 2-4, is an example of a problem that can be solved using constrained search.

In this section, we will show you how to convert this problem into a ConstraintSatisfactionProblem instance using our constraint satisfaction API.

To set up a problem, we first establish a new ConstraintSatisfactionProblem instance. For the Pokemon problem, there are five variables which we pass an an argument in a list: these are the five "questions" that need to be answered.

pokemon_problem = ConstraintSatisfactionProblem(["Q1","Q2","Q3","Q4","Q5"])

Here, we specify the values in each variable's domain:

pokemon_problem.set_domain("Q1",["A","B","C","D","E"])
pokemon_problem.set_domain("Q2",["A","B","C","D","E"])
pokemon_problem.set_domain("Q3",["A","B","C","D","E"])
pokemon_problem.set_domain("Q4",["A","B","C","D","E"])
pokemon_problem.set_domain("Q5",["A","B","C","D","E"])

Next, we set up constraints. Each constraint takes two variable names, and a named binary predicate (constraint function), not a lambda function:

pokemon_problem.add_constraint("Q1","Q4", constraint_different)
pokemon_problem.add_constraint("Q1","Q2", constraint_equal)
pokemon_problem.add_constraint("Q3","Q2", constraint_different)
pokemon_problem.add_constraint("Q3","Q4", constraint_different)
pokemon_problem.add_constraint("Q4","Q5", constraint_equal)

By default, the unassigned_vars list is initialized in alphabetical order.

To specify the order yourself, you can call .set_unassigned_vars_order with an ordered list of the unassigned variables:

# How to set the order of unassigned variables (not actually used for the Pokemon problem)
pokemon_problem.set_unassigned_vars_order(["Q4","Q2","Q3","Q1","Q5"])

For some problems, efficiently re-ordering the variables can make a large difference in performance.

Note that the Pokemon problem is already defined for you in test_problems.py. To get a copy of it, use the method get_pokemon_problem() in lab4.py.

FAQ

Q: I am getting the right output but the wrong number of evaluations

A: Check that, when reducing domains, you are correctly considering the possibility of having multiple different constraints between two variables. (What does it mean if you have two contradictory constraints between two variables?)

Survey

Please answer these questions at the bottom of your lab file:

NAME: What is your name? (string)

COLLABORATORS: Other than 6.034 staff, whom did you work with on this lab? (string, or empty string if you worked alone)

HOW_MANY_HOURS_THIS_LAB_TOOK: Approximately how many hours did you spend on this lab? (number or string)

WHAT_I_FOUND_INTERESTING: Which parts of this lab, if any, did you find interesting? (string)

WHAT_I_FOUND_BORING: Which parts of this lab, if any, did you find boring or tedious? (string)

(optional) SUGGESTIONS: What specific changes would you recommend, if any, to improve this lab for future years? (string)

(We'd ask which parts you find confusing, but if you're confused you should really ask a TA.)

When you're done, run the online tester to submit your code.

Lab 3: Constraint Satisfaction

From 6.034 Wiki

Current revision

Contents

A Working Vocabulary

Part 1: Warm-up

Part 2: Writing a depth-first search solver

Benchmarks

Part 3: Forward checking streamlines search by eliminating impossible assignments

Finding inconsistent values in neighbors

Depth-first constraint solver with forward checking

Part 4: Propagation!

Domain reduction

Propagation through reduced domains

Part 5A: Generic propagation

Part 5B: A generic constraint solver

Part 6: Defining your own constraints

API

Constraint Satisfaction Problems

Constraint objects

Appendix: Setting up a Constraint Satisfaction Problem

FAQ

Survey

Views

Personal tools

Navigation

Search

Toolbox

@@ Line 1: / Line 1: @@
+<!-- {{Unreleased}} -->
+{{Lab_Due|when=Thursday, October 1}}
 __TOC__
-This lab is due by Tuesday, October 27 at 10:00pm.
+{{Get_Code|lab=3}}
-To work on this lab, you will need to get the code, much like you did for the first two labs.
+== A Working Vocabulary ==
-* You can view the files '''and change log''' at: http://web.mit.edu/6.034/www/labs/lab4/
-* Download it as a ZIP file: http://web.mit.edu/6.034/www/labs/lab4/lab4.zip
-* Or, on Athena, <tt>add 6.034</tt> and copy it from <tt>/mit/6.034/www/labs/lab4/</tt>.
+Before beginning, you may want to (re)familiarize yourself with the following terms:
+* '''variable''': something that can receive an assignment value
+* '''value''': something that can be assigned
+* '''domain''': a set of values
+* '''constraint''': a condition that limits domains of possible values
-'''Note:''' If you downloaded the lab files on or before October 15, you will need to download a new version of [http://web.mit.edu/6.034/www/labs/lab4/tester.py tester.py] in order to transmit <tt>ConstraintSatisfactionProblem</tt> objects to the server and pass the online tests.
+== Part 1: Warm-up ==
-<!--Online tests will be made available shortly.  In the meantime, the local tester provides thorough unit tests for each section of the lab.-->
+In this lab, you'll write programs that solve constraint satisfaction problems (CSPs). A CSP consists of variables, assignments, and constraints, and is represented by a <tt>ConstraintSatisfactionProblem</tt> object as described in [[#API | the API]].
-Your answers for this lab belong in the main file <tt>lab4.py</tt>.
+First, we'll get familiarity with CSPs by writing a few helper routines.
-<div id="outline-container-sec-1" class="outline-2">
+* <tt>has_empty_domains(csp)</tt>: returns <tt>True</tt> if the supplied problem has one or more empty domains. Otherwise, returns <tt>False</tt>.
-== Problems ==
+* <tt>check_all_constraints(csp)</tt>: returns <tt>False</tt> if the problem's <b>assigned values</b> violate some constraint. Otherwise, returns <tt>True</tt>.
-<div class="outline-text-2" id="text-1">
-</div><div id="outline-container-sec-1-1" class="outline-3">
-=== Setting up a Constraint Satisfaction Problem ===
-<div class="outline-text-3" id="text-1-1">
-Before beginning, you may want to familiarize yourself with the
+Each function takes in an argument <tt>csp</tt>, which is a [[#Constraint_Satisfaction_Problems|<tt>ConstraintSatisfactionProblem</tt>]] instance.
-following terms: variable, value, domain, constraint, assignment.
-The Pokemon problem from [http://courses.csail.mit.edu/6.034f/Examinations/2012q2.pdf 2012 Quiz 2], pages 2-4, is an example of a problem that can be solved using constrained search:
+== Part 2: Writing a depth-first search solver ==
-This part of the lab will show you how to convert this problem into a
+Now you can use your helper functions to write the constraint solver:
-<code>ConstraintSatisfactionProblem</code> instance using our constraint
-satisfaction API.  The complete documentation is here: [[#API Reference Documentation|API]].
-This problem requires two kinds of constraint: a must-be-equal
+ def solve_constraint_dfs(problem) :
-constraint and a must-not-be-equal constraint. In this lab,
-constraints are defined using functions that take the values of two
-variables as arguments and return True or False based on whether the
-constraint is satisfied or not.
-Here are definitions for the must-be-equal and must-not-be-equal
+This is just like depth-first search as implemented in the search lab, but this time the items in the agenda are partially-solved problems instead of paths. Additionally, for this problem, we will also want to track the number of extensions so we can compare the different strategies for constraint propagation. At the end, instead of returning just a solution, you will return a tuple <tt>(solution, num_extensions)</tt>, where
-constraint functions.  (These are defined in constraint_api.py.)
+* <tt>solution</tt> is the solution to this problem as a dictionary mapping variables to assigned values (see API for details); or <tt>None</tt> if there is no solution to the problem.
+* <tt>num_extensions</tt> is the number of extensions performed during the search. Recall that as before, an extension occurs whenever a problem is <b>removed</b> from the agenda for processing.
- def constraint_equal(a,b) :
+Here is a rough outline of how to proceed:
-    return a == b
- def constraint_different(a,b) :
+# Initialize your agenda and the extension count.
-    return a != b
+# Until the agenda is empty, pop the first problem off the list and increment the extension count.
+# If any variable's domain is empty or if any constraints are violated, the problem is ''unsolvable'' with the current assignments.
+# If none of the constraints have been violated, check whether the problem has any unassigned variables. If not, you've found a complete solution!
+# However, if the problem has some unassigned variables:
+## Take the first unassigned variable off the list using <code>csp.pop_next_unassigned_var()</code>.
+## For each value in the variable's domain, create a new problem with that value assigned to the variable, and add it to a list of new problems. Then, add the new problems to the appropriate end of the agenda.
+# Repeat steps 2 through 6 until a solution is found or the agenda is empty.
-To set up the problem, we first establish a new
+=== Benchmarks ===
-<code>ConstraintSatisfactionProblem</code> instance. There are five variables
-which we pass an an argument in a list — the five questions to be answered.
-<pre class="src src-python" id="example-setup-vars">pokemon_problem = ConstraintSatisfactionProblem(["Q1","Q2","Q3","Q4","Q5"])
+So that we can compare the efficiency of different types of constraint-satisfaction algorithms, we'll compute how many extensions (agenda dequeues) each algorithm requires when solving a particular CSP. Our test problem will be the Pokemon problem from [http://courses.csail.mit.edu/6.034f/Examinations/2012q2.pdf 2012 Quiz 2], pages 2-4.
-</pre>
-Here, we specify the values  in each variable's domain.
+You can solve the Pokemon problem by calling <tt>solve_constraint_dfs(pokemon_problem)</tt> directly. Note that the Pokemon problem is already defined for you in test_problems.py. To get a copy of it, use the method get_pokemon_problem() in lab4.py.
-<pre class="src src-python" id="example-setup-domains">pokemon_problem.set_domain("Q1",["A","B","C","D","E"])
+Please answer the following questions in your lab file:
-pokemon_problem.set_domain("Q2",["A","B","C","D","E"])
-pokemon_problem.set_domain("Q3",["A","B","C","D","E"])
-pokemon_problem.set_domain("Q4",["A","B","C","D","E"])
-pokemon_problem.set_domain("Q5",["A","B","C","D","E"])
-</pre>
-Next, we set up constraints. Each constraint takes two variable names,
+;Question 1
-and a binary predicate.  (For more details on the binary predicate,
+:How many extensions does it take to solve the Pokemon problem with just DFS?
-see the API Reference Documentation.)
-<pre class="src src-python">pokemon_problem.add_constraint("Q1","Q4", constraint_different)
+Put your answer (as an integer) in for <tt>ANSWER_1</tt>.
-pokemon_problem.add_constraint("Q1","Q2", constraint_equal)
+== Part 3: Forward checking streamlines search by eliminating impossible assignments ==
-pokemon_problem.add_constraint("Q3","Q2", constraint_different)
+One problem with the <tt>solve_constraint_dfs</tt> algorithm is that it explores all possible branches of the tree. We can use a trick called forward checking to avoid exploring branches that cannot possibly lead to a solution: each time we assign a value to a variable, we'll eliminate incompatible or ''inconsistent'' values from that variable's neighbors.
-pokemon_problem.add_constraint("Q3","Q4", constraint_different)
-pokemon_problem.add_constraint("Q4","Q5", constraint_equal)
+=== Finding inconsistent values in neighbors ===
-</pre>
-By default, the <tt>unassigned_vars</tt> list is initialized in alphabetical order.
+First, we will write a helper function to eliminate ''inconsistent'' values from a variable's neighbors' domains:
+: Suppose V is a variable with neighbor W.  If W's domain contains a value w which violates a constraint with '''every value in V's domain''', then the assignment W=w can't be part of the solution we're constructing &mdash; we can safely eliminate w from W's domain.
-To specify the order yourself, you can call <tt>.set_unassigned_vars_order</tt> with an ordered list of the unassigned variables:
+(Note that unlike the <tt>check_all_constraints</tt> function above, <tt>eliminate_from_neighbors</tt> checks all combinations of values, and is not restricted to comparing only variables that have assigned values.)
-<pre class="src src-python"># (An alternative, which we won't use here.)
+<!-- :For a given neighbor <tt>n</tt> of a variable <tt>v</tt>, if <tt>n</tt> has a value <tt>nval</tt> that violates a constraint with every value in <tt>v</tt>'s domain, then <tt>nval</tt> is ''inconsistent'' with <tt>n</tt> and <tt>v</tt> and should be removed from <tt>n</tt>'s domain. !-->
-pokemon_problem.set_unassigned_vars_order(["Q4","Q2","Q3","Q1","Q5"])
-</pre>
-(For some problems, efficiently re-ordering the variables can make a large difference in performance.)
+This function should return an alphabetically sorted list of the neighbors whose domains were reduced (i.e. which had values eliminated from their domain), with each neighbor appearing '''at most once''' in the list. If no domains were reduced, return an empty list; if a domain is reduced to size 0, quit and immediately return <tt>None</tt>. This method '''should''' modify the input CSP. Hint: You can remove values from a variable's domain using csp.eliminate(var, val). But don't eliminate values from a variable while iterating over its domain, or Python will get confused!
-We have set up the variables, the domains, and the constraints.
+ def eliminate_from_neighbors(csp, var) :
-The Pokemon problem is defined for you in <tt>test_problems.py</tt>. To get a copy of it, use the method <tt>get_pokemon_problem()</tt> in lab4.py.
+We strongly suggest working out examples on paper to get a feel for how the forward checker should find inconsistent values.
-This problem is now ready to be solved — all we need is a CSP solver.
+To reduce the amount of nested for-loops and to make debugging easier, you may find it helpful to write a small helper function that, for example, takes in two variables V and W, and two values v and w in their respective domains, and checks if there are any constraint violations between V=v and W=w.
-=== Writing a depth-first search solver ===
+=== Depth-first constraint solver with forward checking ===
-Your first task is to write a constraint solver that uses <b>depth-first search</b> to
+Now, we will write our improved CSP solver which uses <tt>eliminate_from_neighbors</tt> above to apply forward checking while searching for variable assignments.
-find a solution.  Start by writing these two helper functions, which may be useful within <code>solve_constraint_dfs</code>:
-  def has_empty_domains(csp) :
+  def solve_constraint_forward_checking(problem) :
-     "Returns True if csp has one or more empty domains, otherwise False"
- def check_all_constraints(csp) :
+The implementation for this function will be very similar to that of <tt>solve_constraint_dfs</tt>, except now the solver must apply forward checking (<tt>eliminate_from_neighbors</tt>) after each assignment, to eliminate incompatible values from the assigned variable's neighbors.
-     """Return False if the problem's assigned values violate some constraint,
-     otherwise True"""
-Each function takes in an argument <tt>csp</tt>, which is a <tt>ConstraintSatisfactionProblem</tt> instance.
+Note that if <tt>eliminate_from_neighbors</tt> eliminates all values from a variable's domain, the problem will be recognized as unsolvable when it is ''next'' popped off the agenda: do not preemptively remove it from consideration.
-Now you can use the helper functions to write the constraint solver:
+Answer the following question in your <tt>lab4.py</tt> file:
-<pre class="src src-python">def solve_constraint_dfs(problem) :
+;Question 2
-</pre>
+:How many extensions does it take to solve the Pokemon problem with forward checking?
-Hint: This is exactly depth-first search as in the search lab, but
+Put your answer (as an integer) in for <tt>ANSWER_2</tt>.
-this time the items in the agenda are partially-solved problems, not
-paths.
-Here is a rough outline of how to proceed:
+== Part 4: Propagation! ==
-<ol class="org-ol">
+Forward checking is a useful tool for checking ahead for inconsistencies and reducing the search space. However, in many situations, it's ideal to prune inconsistent states even faster.
-<li>
-Create an agenda containing only the problem <code>csp</code>. Initialize the extension count to 0.
-</li>
+=== Domain reduction ===
-<li>
-Until the agenda is empty, pop the first problem off the list and increment the extension count by 1.
-</li>
-<li>
-Immediately check if the domain of any of the variables is empty. If any domain is empty, the problem is unsolvable with the current assignments; continue with the next iteration of the loop. Otherwise, begin solving the problem.
-</li>
-<li>
-For that problem, check all the constraints between the variables
-that have been assigned values so far. If any constraint is
-violated, the problem cannot be solved with the given assignments;
-skip the problem and proceed to the next iteration of the
-loop. (The addition of this constraint-checking step is
-where constrained search differs from ordinary search.)
-</li>
+A far-reaching strategy called ''domain reduction'' eliminates incompatible values not just between neighbors, but across all variables in the problem. You can apply domain reduction either ''before search'' (this is what Sudoku players do when they narrow down options before tentatively guessing a value) or ''after assigning each variable during search'' (as a more powerful variation of forward-checking).
-<li>
-If none of the constraints have been violated, check whether the
-problem has any unassigned variables (<code>csp.unassigned_vars</code>). If
-the problem has no unassigned variables, you've found a complete
-solution!  Return the assignments (<code>csp.assigned_values</code>).
-The return type should be a tuple <code>(solution, extension_count)</code>, containing:
+As it turns out, the implementation for both of these are effectively identical:
-# the solution (<code>csp.assigned_values</code>, which is a dictionary mapping variables to assigned values), and
+# Establish a queue. If using domain reduction ''during'' search, this queue should initially contain only the variable that was just assigned. If before search (or if no queue is specified), the queue can contain all variables in the problem.  (Hint: <tt>csp.get_all_variables()</tt> will make a copy of the variables list.)
-# the number of extensions made (the number of problems popped off the agenda).
+# Until the queue is empty, pop the first variable <tt>var</tt> off the queue.
+# Iterate over that <tt>var</tt>'s neighbors: if some neighbor <tt>n</tt> has values that are incompatible with the constraints between <tt>var</tt> and <tt>n</tt>, remove the incompatible values from <tt>n</tt>'s domain. If you reduce a neighbor's domain, add that neighbor to the queue (unless it's already in the queue).
+# If any variable has an empty domain, quit immediately and return <tt>None</tt>.
+# When the queue is empty, domain reduction has finished. Return a list of all variables that were dequeued, in the order they were removed from the queue. Variables may appear in this list multiple times.
-Note: The number of extensions should include the original problem (the root of the search tree),
+Note that when the queue initially contains only the assigned variable, the first step of propagation is just forward checking of the assigned variable's neighbors. "Propagation" occurs as we add more variables to the queue, checking neighbors of neighbors, etc.
-all problems that have children in the tree, any problems that caused search to backtrack, and the final problem (with all variables assigned).
-</li>
-<li>
-If the problem has some unassigned variables:
-<ul class="org-ul">
+You will now implement <tt>domain_reduction</tt>, which applies forward checking (checking for neighboring values' inconsistencies) with propagation through any domains that are reduced.
-<li>Take the first unassigned variable off the list
-(<tt>csp.pop_next_unassigned_var()</tt>).
-</li>
-<li>For each value in the variable's domain (<tt>csp.get_domain(var)</tt>),
-create a new copy of the problem. (<tt>csp_new = csp.copy()</tt>).
-</li>
-<li>Using the copy, assign the value to the variable. (
-<code>csp_new.set_assigned_value(var, val)</code> ).
-</li>
-<li>Collect the list of copies and add the list to the front of the
-agenda (as is appropriate for depth-first search).
-For consistency, keep the list of copies in the same order that the copies were created (the same as the order of values in the variable's domain).
-</li>
-</ul>
-</li>
-<li>
-If the loop has exited because the agenda is empty, the problem is completely unsolvable. Return a tuple comprised of <tt>None</tt> (instead of a solution) and the number of extensions.
-</li>
-</ol>
-=== Domain reduction before search ===
+Recall that domain reduction utilizes a queue to keep track of the variables whose neighbors should be explored for inconsistent domain values. If you are not explicitly provided a queue from the caller, your queue should start out with all of the problem's variables in it, in their default order.
-Domain reduction is a strategy for eliminating impossible values in
+When doing domain reduction, you should keep track of the order in which variables were dequeued; the function should return this ordered list of variables that were dequeued.
-advance to cut down on the amount of search you have to do.
-First, write a helper function to eliminate values from a variable's neighbors' domains.
+If at any point in the algorithm a domain becomes empty, immediately return <tt>None</tt>.
-This function should return an alphabetically sorted list of the neighbors
-whose domains were reduced, with each neighbor appearing AT MOST ONCE in the list (the list should contain no duplicates).
-If no domains are reduced, return an empty list;
-if a domain is reduced to size 0 (no values left in domain), quit and immediately return None.
-This method should modify the input csp.
-Hint: <tt>csp.constraints_between</tt> may help.
+ def domain_reduction(csp, queue=None) :
- def eliminate_from_neighbors(csp, var) :
+This method '''should''' modify the input CSP.
-Next, you will need to write a domain-reduction algorithm.
+Hint: You can remove values from a variable's domain using <tt>csp.eliminate(var, val)</tt>. But '''don't''' eliminate values from a variable while iterating over its domain, or Python will get confused!
-<pre class="src src-python">def domain_reduction(csp, queue=None) :
+Answer the following question in your <tt>lab4.py</tt> file:
-</pre>
-Here is a rough description of the domain reduction algorithm:
+;Question 3
+:How many extensions does it take to solve the Pokemon problem with DFS (no forward checking) if you do domain reduction before solving it?
-<ol class="org-ol">
+Put your answer (as an integer) in for <tt>ANSWER_3</tt>.
-<li>Establish a queue — if the optional argument <code>queue</code> was passed
-as an argument, use that as your queue. Otherwise, <code>queue</code> should
-consist of all the variables in the problem in their default order. (Hint: <tt>csp.get_all_variables()</tt> will make a copy of the variables list.)
-</li>
-<li>Until the queue is empty, pop the first variable off the queue. <!-- todo should 'pop' be 'dequeue'?  does it matter? (probably not.) -->
-</li>
-<li>Iterate over the variable's constraints: if a neighbor has a value
-that violates a constraint with <i>every</i> value in the first variable's
-domain, remove the neighbor's value and add the neighbor to the
-queue, unless the neighbor is already in the queue.
-Add neighbors to the queue in alphabetical order (this will probably be the same as the default order).
-<!--<sup><a id="fnr.1" class="footref" href="#fn.1">1</a></sup>. [todo: convert 'footref' tags into links to the Footnotes section]-->
-</li>
-<li>If any variable has an empty domain, quit immediately and return
-None.
-</li>
-<li>When the queue is empty, domain reduction has finished. As a
-side-effect, please return a list of all variables that were dequeued,
-in the order they were removed from the queue.  Variables may appear in this
-list multiple times.
-</li>
-</ol>
-This method should modify the input csp.
+=== Propagation through reduced domains ===
-Hint: You can remove values from a variable's domain using <code>csp.eliminate(var, val)</code>.
+Now we'll see how we can take advantage of domain reduction during the search procedure itself.
-(But don't eliminate values from a variable while iterating over its domain, or Python will get confused.)
-<h4 id="sec-1-3-1"> Benchmarks</h4>
+When used during search, domain reduction makes use of the assignments you've made to progressively reduce the search space. The result is a new, faster, CSP solution method: propagation through reduced domains.  After each assignment, propagation through reduced domains uses the domain_reduction subroutine to "propagate" the consequences of the assignment: to neighbors, then to neighbors of neighbors, and so on.
-You can solve the pokemon problem by calling
+ def solve_constraint_propagate_reduced_domains(problem) :
-<code>solve_constraint_dfs(pokemon_problem)</code> directly. Using domain
-reduction first, however, should make search faster. Please answer the
-following questions in your lab file:
+Note that if <tt>domain_reduction</tt> eliminates all values from a variable's domain, the problem will be recognized as unsolvable when it is ''next'' popped off the agenda: do not preemptively remove it from consideration.
-QUESTION 1: How many extensions does it take to solve the Pokemon problem
+Debugging hint: be sure to look at the return types of functions that you call!
-with dfs if you DON'T use domain reduction before solving it?
-Hint: Use get_pokemon_problem() to get a new copy of the Pokemon problem
+Answer the following question in your <tt>lab4.py</tt> file:
-each time you want to solve it with a different search method.
- ANSWER_1 = None
+;Question 4
+:How many extensions does it take to solve the Pokemon problem with forward checking and propagation through reduced domains? (Don't use domain reduction before solving it.)
+Put your answer (as an integer) in for <tt>ANSWER_4</tt>.
-QUESTION 2: How many extensions does it take to solve the Pokemon problem
+== Part 5A: Generic propagation ==
-with dfs if you DO use domain reduction before solving it?
- ANSWER_2 = None
+The <tt>domain_reduction</tt> procedure is comprehensive, but expensive: it eliminates as many values as possible, but it continually adds more variables to the queue. As a result, it is an effective algorithm to use ''before'' solving a constraint satisfaction problem, but is often too expensive to call repeatedly during search.
-=== Propagation through reduced domains ===
+Instead of comprehensively reducing all the domains in a problem, as <tt>domain_reduction</tt> does, you can instead reduce only ''some'' of the domains. This idea underlies ''propagation through singleton domains'' — a reduction algorithm which does not detect as many dead ends, but which is significantly faster.
-Domain reduction can be used not only before search, but also <i>during</i>
-search. When used during search, domain reduction can make use of the
-assignments you've made to progressively reduce the search space. The
-result is a new, faster, csp solution method: propagation through
-reduced domains.
-<pre class="src src-python">def solve_constraint_propagate_reduced_domains(problem) :
+Instead of again patterning our propagation-through-singleton-domains algorithm off of <tt>domain_reduction</tt>, we'll write a fully general propagation algorithm called <tt>propagate</tt> that encapsulates all three checking strategies we've seen: forward checking, propagation through all reduced domains, and propagation through singleton domains.
-    raise NotImplementedError
-</pre>
-You can reuse most of your code from your solve_constraint_dfs
+The function <code>propagate</code> will be similar to the propagation algorithms you've already defined. The difference is that it will take an argument <tt>enqueue_condition_fn</tt>, a function that takes a problem and a variable, and outputs whether the variable should be added to the propagation queue.
-algorithm, as domain reduction adds only a single conditional
-statement: every time you assign a value to a variable, you must call
-your <code>domain_reduction</code> function with a list containing just the assigned variable as an argument. (This will look for and eliminate incompatible values from variables' domains.) Then add the problem to the agenda as usual.
-(If <code>domain_reduction</code> eliminates all values from a variable's domain, the problem will be recognized as unsolvable when it is next popped off the agenda.)
-Note: domain_reduction should remove values from the assigned variable's ''neighbors' ''domains, not from the variable's domain.
+ def propagate(enqueue_condition_fn, csp, queue = None) :
-Answer the following question in your lab4.py file:
+Propagation through singletons is like propagation through reduced domains, except that variables must pass a test in order to be added to the queue:
+:In propagation through singleton domains, you only append a variable to the queue if it has exactly one value left in its domain.
-QUESTION 3: How many extensions does it take to solve the Pokemon problem
+'''Common misconception''': Please note that propagation '''never''' ''assigns'' values to variables; it only ''eliminates'' values. There is a distinction between variables with one value in their domain, and assigned variables: a variable can have one value in its domain without any value being assigned yet.
-with propagation through reduced domains? (Don't use domain reduction before solving it.)
- ANSWER_3 = None
+----
-=== Propagation through singleton domains ===
+As a review, propagation eliminates incompatible options from neighbors of variables in the queue. When used during search, the propagation queue initially contains only the just-assigned variable. The three enqueueing conditions we've seen are:
-The <code>reduce_domains</code> procedure is comprehensive, but expensive: it
+# ''forward checking'': never adds other variables to the queue
-eliminates as many values as possible, but it continually adds more
+# ''propagation through singleton domains'': adds a neighboring variable to the queue if its domain has exactly one value in it
-variables to the queue. As a result, it is an effective algorithm to
+# ''domain reduction / propagation through reduced domains'': adds a neighboring variable to the queue if its domain has been reduced in size
-use <i>before</i> solving a constraint satisfaction problem, but is often
-too expensive to call repeatedly during search.
-Instead of comprehensively reducing all the domains in a problem, as
+Write functions that represent the enqueueing conditions (predicates) for each of these. Each predicate function below takes in a CSP and the variable in question, returning <tt>True</tt> if that variable should be added to the propagation queue, otherwise <tt>False</tt>.
-<code>reduce_domains</code> does, you can instead reduce only <i>some</i> of the
+<pre>
-domains. This results in <i>propagation through singleton domains</i> — a
+ def condition_domain_reduction(csp, var) :
-reduction algorithm which does not detect as many dead ends, but which
-is significantly faster.
-<pre class="src src-python">def domain_reduction_singleton_domains(csp, queue=None) :
+ def condition_singleton(csp, var) :
+ def condition_forward_checking(csp, var) :
 </pre>
-Propagation through singletons is like propagation through reduced
+== Part 5B: A generic constraint solver ==
-domains, except that variables must pass a test in order to be added to the queue:
-<blockquote>
+Now, you can use <tt>propagate</tt> to write a generic constraint solver. Write an algorithm that can solve a problem using any enqueueing strategy. As a special case, if the <tt>enqueue_condition</tt> is <code>None</code>, default to ordinary dfs instead --- don't eliminate options from neighbors (don't use any forward checking or propagation) at all.
-In propagation through singleton domains, you only append a variable
-to the queue if it has exactly one value left in its domain.
-</blockquote>
-<b>Common misconception</b>: Please note that propagation never <i>assigns</i>
+ def solve_constraint_generic(problem, enqueue_condition=None) :
-values to variables; it only eliminates values. There is a distinction
-between variables with one value in their domain, and assigned
-variables: a variable can have one value in its domain without any
-value being assigned yet.
-Now, you can use <tt>domain_reduction_singleton_domains</tt> to write a constraint solver that propagates through singleton domains:
+Answer the following question in your <tt>lab4.py</tt> file:
- def solve_constraint_propagate_singleton_domains(problem) :
+;Question 5
+:How many extensions does it take to solve the Pokemon problem with forward checking and propagation through singleton domains? (Don't use domain reduction before solving it.)
+Put your answer (as an integer) in for <tt>ANSWER_5</tt>.
-Answer the following question in your lab4.py file:
+== Part 6: Defining your own constraints ==
-QUESTION 4: How many extensions does it take to solve the Pokemon problem
+In this section, you will create some constraint functions yourself.
-with propagation through singleton domains? (Don't use domain reduction before solving it.)
- ANSWER_4 = None
+Assuming <tt>m</tt> and <tt>n</tt> are integers, write a function that returns <tt>True</tt> if <tt>m</tt> and <tt>n</tt> are adjacent values (i.e. if they differ by exactly one) and <tt>False</tt> otherwise.
-=== Forward checking ===
+ def constraint_adjacent(m, n) :
-Forward checking is even more restrictive than propagation through
-singletons: it <i>never</i> adds variables to the queue. <!--- TODO: what is this (dxh): (Later in this
-lab, we will perform tests to see which propagation algorithm
-performs best in terms of tradeoffs between performance and
-comprehensiveness.) !-->
-Instead of patterning our forward checking algorithm off of
+Also write one for being non-adjacent.
-<code>domain_reduction</code> again, we'll write a fully general algorithm called
-<code>propagate</code> that encapsulates all three propagation strategies we've
-seen: propagation through reduced domains, propagation through
-singletons, and forward checking.
-The function <code>propagate</code> will be similar to the propagation algorithms
+ def constraint_not_adjacent(m, n) :
-you've already defined. The difference is that it will take an
-argument <code>enqueue_condition_fn</code> which is the test that variables must
-pass in order to be added to the queue: before <code>propagate</code> adds a
-variable to the queue, it should call <code>enqueue_condition_fn(csp, var)</code>
-. If the function returns <code>True</code>, it should add the variable to the
-queue. Otherwise, it shouldn't.
-<pre class="src src-python">def propagate(enqueue_condition_fn, csp, queue = None) :
+The following example shows how you build a constraint object that requires two variables — call them A and B — to be different.
-    raise NotImplementedError
-</pre>
-To review, the three enqueueing conditions we've seen are:
+ example_constraint = Constraint("A","B", constraint_different)
-<dl class="org-dl">
+Some constraint problems include a constraint that requires all of the variables to be different from one another. It can be tedious to list all of the pairwise constraints by hand, so we won't. Instead, write a function that takes a list of variables and returns a list containing, for each pair of variables, a constraint object requiring the variables to be different from each other. (You can model the constraints on the example above.) Note that for this ''particular'' constraint (the must-be-different constraint), order does NOT matter, because inequality is a symmetric relation. Hence, in you should only have ''one'' constraint between each pair of variables (e.g. have
-<dt>domain reduction<dd>always add the variable to the queue
+a constraint between A and B, '''OR''' have a constraint between B and A, but not both).
-<dt>singleton propagation<dd>add the variable if its domain has exactly
-one value in it.
-<dt>forward checking<dd>never add the variable to the queue.
-</dl>
-Write functions for each of these tests. Hint: some of these
+ def all_different(variables) :
-functions may ignore some of their arguments.
-<pre class="src src-python">def condition_domain_reduction(csp, var) :
+Note: You should only use constraint functions that have already been defined. Don't try to create a new constraint function and use it in this function, because our tester will get confused.
-def condition_singleton(csp, var) :
+<!-- Future ideas:
+=== Resource allocation ===
+Find out what size domain you need using binary search.
-def condition_forward_checking(csp, var) :
+=== Counting solutions ===
-</pre>
+Use "yield" to count the number of solutions.
-<!-- todo maybe uncomment
+<h4 id="sec-1-7-1"> Comparing propagation strategies</h4>
-And thus we can define:
+Including the trouble with texas.
-<pre class="src src-python">domain_reduction_forward_checking =
+<h4 id="sec-1-7-2"> Variable re-ordering</h4>
-lambda csp, queue=None: propagate(condition_forward_checking, csp, queue)
+How should you order your variables in the agenda? e.g.
-</pre>
+<ul class="org-ul">
+<li>most constrained variables first
+</li>
+<li>smallest domains first
+</li>
+</ul>
+Let's find out empirically.
 -->
-=== A generic CSP solver ===
+== API ==
-Write an algorithm that can solve a problem using any enqueueing
-strategy. As a special case, if the enqueue_condition is <code>None</code>, don't
-use any propagation at all (i.e. the algorithm should perform only
-depth-first search in this case.)
-To propagate a variable, call <tt>propagate</tt> with the specified <tt>enqueue_condition</tt>.
-<pre class="src src-python">def solve_constraint_generic(problem, enqueue_condition=None) :
+In this lab, we provide an API for representing and manipulating partial solutions to constraint satisfaction problems.
-</pre>
+=== Constraint Satisfaction Problems ===
-Answer the following question in your lab4.py file:
+A <tt>ConstraintSatisfactionProblem</tt> is an object representing a partially solved constraint satisfaction problem. Its fields are:
-QUESTION 5: How many extensions does it take to solve the Pokemon problem
+;<tt>variables</tt>
-with DFS and forward checking, but no propagation? (Don't use domain reduction before solving it.)
+:A list containing the names of all the variables in the problem, in alphabetical order.
- ANSWER_5 = None
+;<tt>domains</tt>
+:A dictionary associating each variable in the problem with its list of remaining values.
-=== Solving your own CSPs ===
+;<tt>assignments</tt>
-<h4 id="sec-1-8-1"> Defining new constraints</h4>
+:A dictionary. Each variable that has already been assigned a value is associated with that value here. When the problem is entirely solved, <tt>assignments</tt> contains the solution.
-Assuming m and n are integers, write a function that returns True if
+;<tt>unassigned_vars</tt>
-m and n are adjacent values (i.e. if they differ by exactly one) and
+:An ordered list of all the variables that still need to have a value assigned to them.
-False otherwise.
-<pre class="src src-python">def constraint_adjacent(m, n) :
+;<tt>constraints</tt>
-</pre>
+:A list of the constraints between the variables in the problem. Each constraint is a <tt>Constraint</tt> object.
-Also write one for being non-adjacent. There are trivial ways to do
+Note: While you may ''read'' any of the above variables, you should probably not modify them directly;<tt> instead, you should use the following API methods:</tt>
-it; feel free to call <code>constraint_adjacent</code>.
-<pre class="src src-python">def constraint_not_adjacent(m, n) :
+;<tt>get_domain(var)</tt>
-</pre>
+:Returns the list of values in the variable's domain.
-The following example shows how you build a constraint object that
+;<tt>set_domain(var, domain)</tt>
-requires two variables — call them A and B — to be different.
+:Sets the domain of the variable to the specified list of values, sorted alphabetically/numerically.
-<pre class="src src-python">example_constraint = Constraint("A","B", constraint_different)
+;<tt>set_all_domains(domains_dict)</tt>
-</pre>
+:Sets the <tt>domains</tt> field to the specified dictionary. Does not sort domains.
-Some constraint problems include a constraint that requires all of
+;<tt>get_all_variables()</tt>
-the variables to be different from one another. It can be tedious to
+:Returns a list of all the variables in the problem.
-list all of the pairwise constraints by hand, so we won't. Instead, write a
-function that takes a list of variables and returns a list
-containing, for each pair of variables, a constraint object
-requiring the variables to be different from each other. (You can model the
-constraints on the example above.)  Note that order does NOT matter,
-so should only have one constraint between each pair of variables (ie (A,B) OR
-(B,A)).
-<pre class="src src-python">def all_different(variables) :
+;<tt>get_all_constraints()</tt>
-</pre>
+:Returns a list of all the [[#Constraint_objects | constraints]] in the problem.
-<h4 id="sec-1-8-2"> Defining a new problem: Moose problem (OPTIONAL)</h4>
+;<tt>pop_next_unassigned_var()</tt>
-If you want to try out your new constraints and CSP solver, you may design and solve a constraint satisfaction problem for the Moose Problem from 2008 Quiz 2.  You are of course welcome to implement additional problems; some of our favorites include the Time Travelers problem (2009 Quiz 2) and the Zoo problem (2011 Quiz 2).
+:Returns first unassigned variable, or <tt>None</tt> if all variables are assigned. Modifies <tt>unassigned_vars</tt> list.
-The moose problem can be found on [http://courses.csail.mit.edu/6.034f/Examinations/2008q2.pdf 2008 Quiz 2], page 4.  Use people as variables and seats as values.
+;<tt>set_unassigned_vars_order(unassigned_vars_ordered)</tt>
+:Given an ordered list of unassigned variables, sets <tt>unassigned_vars</tt>. (By default, the <tt>unassigned_vars</tt> list is initialized in alphabetical order.)
-Note: You will need to make a modified version of the constraint_adjacent function above to account for the table being round. (ie, seats 1 and 6 are adjacent, even though the numbers are far apart)
+;<tt>eliminate(var, val)</tt>
+:Removes the value from the variable's domain, returning <tt>True</tt> if the value was found in the domain, otherwise <tt>False</tt> if the value wasn't found.
-Interpret constraints that only mention one person (e.g. "McCain insists on sitting in seat 1") as an indication that you should
+;<tt>get_assignment(var)</tt>
-leave out those values from the variable's domain; you won't
+:If the variable has been assigned a value, retrieve it. Returns <tt>None</tt> if the variable hasn't been assigned yet.
-need to make a Constraint to represent those kinds of constraints.
-Establish the domains of the variables. Each variable will have a
+;<tt>set_assignment(var, val)</tt>
-domain which is a subset of [1,2,3,4,5,6].
+:Sets the assigned value of the variable to <tt>val</tt>, returning a modified copy of the constraint satisfaction problem. Throws an error if <tt>val</tt> is not in the domain of <tt>var</tt>, or if <tt>var</tt> has already been assigned a value. For convenience, also modifies the variable's domain to contain only the assigned value.
-Establish the constraints. Remember the constraint that says that
+;<tt>constraints_between(var1, var2)</tt>
-all of the variables must be assigned a different value. You may
+:Returns a list of all [[#Constraint_objects | constraints]] between <tt>var1</tt> and <tt>var2</tt>. Arguments that are <tt>None</tt> will match anything: for example, <tt>constraints_between('X',None)</tt> will return all constraints involving <tt>X</tt> and any other variable, and <tt>constraints_between(None, None)</tt> will return all of the constraints in the problem.
-want to use <code>csp.add_constraints(list)</code>.
+:Note: For your convenience, the constraints returned will always be altered to match the order of the arguments you passed to this method. For example, <tt>csp.constraints_between(None, 'Y')</tt> will return a list of all constraints involving <tt>'Y'</tt> — and the constraints will be altered so that <tt>'Y'</tt> is their ''second'' variable (<tt>var2</tt>) in every case.
-To run local tests on your Moose Problem, set the boolean flag TEST_MOOSE_PROBLEM to True in lab4.py.
+;<tt>get_neighbors(var)</tt>
+:Returns a list of all the variables that share constraints with the given variable, ordered alphabetically.
-<!-- Future ideas:
+;<tt>add_constraint(var1, var2, constraint_fn)</tt>
-=== Resource allocation ===
+:Given two variables and a function to act as a constraint between them, creates a [[#Constraint_objects | <tt>Constraint</tt> object]] and adds it to the <tt>constraints</tt> list. The function <tt>constraint_fn</tt> must be a binary predicate function: it takes two arguments (a value for the first variable, and a value for the second variable) and returns <tt>True</tt> if the values satisfy the constraint, or <tt>False</tt> otherwise.
-Find out what size domain you need using binary search.
-=== Counting solutions ===
+;<tt>add_constraints(list_of_constraints)</tt>
-Use "yield" to count the number of solutions.
+:Add a list of [[#Constraint_objects | <tt>Constraint</tt> objects]] to the <tt>constraints</tt> list. Useful for when you want to add several constraints to the problem at once, rather than one at a time using <code>.add_constraint</code>.
-<h4 id="sec-1-7-1"> Comparing propagation strategies</h4>
+;<tt>copy()</tt>
-Including the trouble with texas.
+:Return a (deep) copy of this constraint satisfaction problem. This method is particularly useful because you will want to make a copy of the CSP every time you assign a value to a variable.
-<h4 id="sec-1-7-2"> Variable re-ordering</h4>
+=== Constraint objects ===
-How should you order your variables in the agenda? e.g.
-<ul class="org-ul">
-<li>most constrained variables first
-</li>
-<li>smallest domains first
-</li>
-</ul>
-Let's find out empirically.
+A <tt>Constraint</tt> is a fairly basic object representing a constraint between two variables. A <tt>Constraint</tt> object has three fields:
--->
-== API Reference Documentation ==
+;<tt>var1</tt>
-In this lab, we provide an API for representing and
+:The first variable associated with this constraint
-manipulating partial solutions to constraint satisfaction problems.
-<!--
-A <tt>ConstraintSatisfactionProblem</tt> object encodes the state of a partially-solved problem; these are the nodes in your search tree when you are performing constrained search by hand.
-!-->
-=== Constraint Satisfaction Problems ===
+;<tt>var2</tt>
-<div class="outline-text-3" id="text-2-1">
+:The second variable associated with this constraint
-A <code>ConstraintSatisfactionProblem</code> is an object representing a
+;<tt>constraint_fn</tt>
-partially solved constraint satisfaction problem. Its fields are:
+:A function that takes in two arguments, returning <tt>True</tt> or <tt>False</tt> depending on whether or not the given constraint is satisfied by the two arguments. For example,
-<ul>
+:* <tt>constraint_equal(a, b)</tt> is a function requiring that <tt>a</tt> and <tt>b</tt> are equal.
-<li>'''.variables''':
+:* <tt>constraint_different(a, b)</tt> is a function requiring that <tt>a</tt> and <tt>b</tt> are not equal.
-<span></span>A list containing the names of all the variables in the problem, in alphabetical order.
+:These two functions are already defined in <tt>constraint_api.py</tt>, and can be accessed directly from <tt>lab4.py</tt>.
-</li>
-<li>'''.domains''': A dictionary associating each variable in the problem with
-its list of remaining values.<!--<sup><a id="fnr.2" class="footref" href="#fn.2">2</a></sup> todo-->
-</li>
-<li>'''.assigned_values''': A dictionary. Each variable that has already been
-assigned a value is associated with that value
-here. When the problem is entirely solved,
-<code>assigned_values</code> contains the solution.
-</li>
-<li>'''.unassigned_vars''': An ordered list of all the variables that still
-need to have a value assigned to them.
-</li>
-<li>'''.constraints''': A list of the constraints between the variables in
-the problem. Each constraint is a <code>Constraint</code>
-object; <code>Constraint</code> objects are described in the
-next section.
-</li>
-</ul>
-Note: While you may <i>read</i> any of the above variables, you
+----
-should probably not modify them directly; instead, you should use the
-following API methods. (Below, <code>csp</code> stands for some
-<code>ConstraintSatisfactionProblem</code> instance that you want to manipulate.)
-<ul>
+A <tt>Constraint</tt> object has just one method associated with it:
-<li><tt>'''csp'''.get_domain(var)</tt> <span></span>Returns the list of values in the variable's
-domain.
-</li>
-<li><tt>'''csp'''.set_domain(var, domain)</tt> <span></span>Sets the domain of the variable to the
-specified list of values, sorted alphabetically/numerically.
-</li>
-<li><tt>'''csp'''.set_all_domains(domains_dict)</tt> <span></span>Sets the .domains attribute to the specified dictionary.  Does not sort domains.
-</li>
-<li><tt>'''csp'''.get_all_variables()</tt> <span></span>Returns a list of all the variables in the problem.
-</li>
-<li><tt>'''csp'''.get_all_constraints()</tt> <span></span>Returns a list of all the constraints in the problem.
-</li>
-<li><tt>'''csp'''.pop_next_unassigned_var()</tt> <span></span>Returns first unassigned variable, or None if all variables are assigned.  Modifies unassigned_vars list.
-</li>
-<li><tt>'''csp'''.set_unassigned_vars_order()</tt> <span></span>Given an ordered list of unassigned variables, sets the list of unassigned vars.  (By default, the unassigned_vars list is initialized in alphabetical order.)
-</li>
-<li><tt>'''csp'''.eliminate(var, val)</tt> <span></span>Removes the value from the variable's
-domain. <span style="background:#ff0">Note</span> that as a helpful side-effect, this function returns <code>True</code> if the value was found in the
-domain, or <code>False</code> if the value wasn't found.
-</li>
-</ul>
-<ul>
+;<tt>check(val1, val2)</tt>
-<li><tt>'''csp'''.get_assigned_value(var)</tt> <span></span>If the variable has been assigned a
+:Applies this object's <tt>constraint_fn</tt> to two ''values'' ('''not''' variables), returning <tt>True</tt> if the values satisfy the constraint, or <tt>False</tt> otherwise.
-value, retrieve it. Returns <code>None</code> if the variable hasn't been
-assigned yet<!--<sup><a id="fnr.3" class="footref" href="#fn.3">3</a></sup> todo -->.
-</li>
-<li><tt>'''csp'''.set_assigned_value(var, val)</tt> <span></span>Sets the assigned value of the
-variable to <tt>val</tt>, <em>returning a modified copy of the constraint satisfaction
-problem</em>. Throws an error if <tt>val</tt> is not in the domain of the
-variable, or if var has already been assigned a value. For convenience, also modifies the variable's domain to contain only the assigned value.
-</li>
-</ul>
-<ul>
+'''Note:''' Due to certain limitations in our tester, a <code>Constraint</code> object constructor must take a '''named''' <code>constraint_fn</code> as an argument, '''NOT''' a lambda function.
-<li><tt>'''csp'''.constraints_between(var1, var2)</tt> <span></span>
-Returns a list of all the
-constraints between var1 and var2. Arguments that are <code>None</code> will
-match anything: for example, <code>constraints_between('X',None)</code> will
-return all constraints involving <code>X</code> any any other variable, and
-<code>constraints_between(None, None)</code> will return all of the
-constraints in the problem.
-Note: For your convenience, the constraints returned
+== Appendix: Setting up a Constraint Satisfaction Problem ==
-will always be altered to match the order of the arguments you
-passed to this method. For example,
-<code>csp.constraints_between(None, 'Y')</code> will return a list of all
-constraints involving 'Y' — and the constraints will be altered
-so that 'Y' is their <i>second</i> variable (<code>var2</code>) in every case.
-</li>
+The Pokemon problem from [http://courses.csail.mit.edu/6.034f/Examinations/2012q2.pdf 2012 Quiz 2], pages 2-4, is an example of a problem that can be solved using constrained search.
-<li><tt>'''csp'''.add_constraint(var1, var2, constraint_fn)</tt> <span></span>
+In this section, we will show you how to convert this problem into a <tt>ConstraintSatisfactionProblem</tt> instance using our [[#API|constraint satisfaction API]].
-Given two variables and a function to act as a constraint between them, creates a
-<code>Constraint</code> object and adds it to the list
-<code>csp.constraints</code>.
-The function <code>constraint_fn</code> must take two arguments — a value for
+To set up a problem, we first establish a new <code>ConstraintSatisfactionProblem</code> instance. For the Pokemon problem, there are five variables which we pass an an argument in a list: these are the five "questions" that need to be answered.
-the first variable, and a value for the second variable — and
-return <code>True</code> if the values satisfy the constraint, or <code>False</code>
-otherwise.
-</li>
+ pokemon_problem = ConstraintSatisfactionProblem(["Q1","Q2","Q3","Q4","Q5"])
-<li><tt>'''csp'''.add_constraints(constraint_list)</tt> <span></span>
-Add a list of <code>Constraint</code>
-objects to the list <code>csp.constraints</code>. Useful for when you want
-to add several constraints to the problem at once, rather than
-one at a time using <code>csp.add_constraint</code>.
-</li>
+Here, we specify the values in each variable's domain:
-</ul>
-<ul>
+<pre>
-<li><tt>'''csp'''.copy()</tt> <span></span>Return a (deep) copy of this constraint satisfaction
+pokemon_problem.set_domain("Q1",["A","B","C","D","E"])
-problem. This method is particularly useful because
+pokemon_problem.set_domain("Q2",["A","B","C","D","E"])
-you will want to make a copy of the <code>csp</code> every time
+pokemon_problem.set_domain("Q3",["A","B","C","D","E"])
-you assign a value to a variable.
+pokemon_problem.set_domain("Q4",["A","B","C","D","E"])
-</li>
+pokemon_problem.set_domain("Q5",["A","B","C","D","E"])
-</ul>
+</pre>
-=== Constraint objects ===
+Next, we set up constraints. Each constraint takes two variable names, and a ''named'' [[#API|binary predicate]] (constraint function), not a lambda function:
-A <code>Constraint</code> is a fairly basic object. It has three variables—
-<code>var1</code>, <code>var2</code>, and <code>constraint_fn</code> — and one method, <code>check(val1,
-val2)</code>.
-<code>constraint.check(val1, val2)</code> simply applies the <code>Constraint</code>'s
+<pre>
-constraint function to the two arguments, returning <code>True</code> if the
+pokemon_problem.add_constraint("Q1","Q4", constraint_different)
-values satisfy the constraint, otherwise <code>False</code>.
+pokemon_problem.add_constraint("Q1","Q2", constraint_equal)
+pokemon_problem.add_constraint("Q3","Q2", constraint_different)
+pokemon_problem.add_constraint("Q3","Q4", constraint_different)
+pokemon_problem.add_constraint("Q4","Q5", constraint_equal)
+</pre>
+By default, the <tt>unassigned_vars</tt> list is initialized in alphabetical order.
+To specify the order yourself, you can call <tt>.set_unassigned_vars_order</tt> with an ordered list of the unassigned variables:
+<pre>
+# How to set the order of unassigned variables (not actually used for the Pokemon problem)
+pokemon_problem.set_unassigned_vars_order(["Q4","Q2","Q3","Q1","Q5"])
+</pre>
+For some problems, efficiently re-ordering the variables can make a large difference in performance.
+----
+Note that the Pokemon problem is already defined for you in <tt>test_problems.py</tt>. To get a copy of it, use the method <tt>get_pokemon_problem()</tt> in <tt>lab4.py</tt>.
 <!--=== <span class="todo TODO">TODO</span> Worked examples === -->
@@ Line 614: / Line 409: @@
 -->
-== Survey ==
+== FAQ ==
-Please answer these questions at the bottom of your <tt>lab3.py</tt> file:
-* <tt>NAME</tt>: What is your name? (string)
-* <tt>COLLABORATORS</tt>: Other than 6.034 staff, whom did you work with on this lab? (string, or empty string if you worked alone)
-* <tt>HOW_MANY_HOURS_THIS_LAB_TOOK</tt>: Approximately how many hours did you spend on this lab? (number or string)
-* <tt>WHAT_I_FOUND_INTERESTING</tt>: Which parts of this lab, if any, did you find interesting? (string)
-* <tt>WHAT_I_FOUND_BORING</tt>: Which parts of this lab, if any, did you find boring or tedious? (string)
-* (optional) <tt>SUGGESTIONS</tt>: What specific changes would you recommend, if any, to improve this lab for future years? (string)
+'''Q''': I am getting the right output but the wrong number of evaluations
+'''A''': Check that, when reducing domains, you are correctly considering the possibility of having multiple different constraints between two variables. (What does it mean if you have two contradictory constraints between two variables?)
-(We'd ask which parts you find confusing, but if you're confused you should really ask a TA.)
-When you're done, run the online tester to submit your code.  <!--(The online tester for Lab 4 will be made available by ___.)-->
+{{Survey}}