We've created a LaTeX template here for you to use that contains the prompts for each question.


What courses should you take in a given quarter? Answering this question requires balancing your interests, satisfying prerequisite chains, graduation requirements, availability of courses; this can be a complex tedious process. In this assignment, you will write a program that does automatic course scheduling for you based on your preferences and constraints. The program will cast the course scheduling problem (CSP) as a constraint satisfaction problem (CSP) and then use backtracking search to solve that CSP to give you your optimal course schedule.

You will first get yourself familiar with the basics of CSPs in Problem 0. In Problem 1, you will implement a heuristic you learned from lecture that will make CSP solving much faster. Lastly, in Problem 2, you will create the course scheduling CSP and solve it using the code from previous parts.

Problem 0: CSP basics

  1. Let's create a CSP. Suppose you have $n$ light bulbs, where each light bulb $i = 1, \dots, n$ is initially off. You also have $m$ buttons which control the lights. For each button $j = 1, \dots, m$, we know the subset $T_j \subseteq \{ 1, \dots, n \}$ of light bulbs that it controls. When button $j$ is pressed, it toggles the state of each light bulb in $T_j$ (for example, if $3 \in T_j$ and light bulb 3 is off, then after the button is pressed, light bulb 3 will be on, and vice versa). If multiple buttons—say $Z$ buttons—controlling the same light bulb are pressed, then that light bulb will be turned on if $Z$ is odd, or it will be turned off if $Z$ is even.

    Your goal is to turn on all the light bulbs by pressing a subset of the buttons. Construct a CSP to solve this problem. Your CSP should have $m$ variables and $n$ constraints. For this problem only, you can use $m$-ary constraints: constraints that can be functions of up to $m$ variables. Describe your CSP precisely and concisely. You need to specify the variables with their domain, and the constraints with their scope and expression. Make sure to include $T_j$ in your answer.

    Hint: If stuck, take a look at parts (b) and (c) of this problem to see how you could define the constraints using a boolean operator.

    A clear description of your solution CSP including $m$ variables and their domains, and $n$ constraints.

  2. Now, let's consider a simple CSP with 3 variables and 2 binary factors:

    0a - fancy CSP

    where $X_1,X_2,X_3 \in \{0,1\}$ and $t_1, t_2$ are XOR functions (that is $t_1(X) = x_1 \bigoplus x_2$ and $t_2(X) = x_2 \bigoplus x_3$).
    1. What are the consistent assignments for this CSP?
    2. Let's use backtracking search to solve the CSP without using any heuristics(MCV, LCV, forward checking, AC-3). The Backtrack() algorithm as defined in the lectures is a recursive algorithm, where new instances of Backtrack() are called within parent instances of Backtrack().

      In this problem, we will ask you to produce the call stack for a specific call to Backtrack(). A call stack is just a diagram tracing out every recursive call. For our purposes, for each call to Backtrack() you should specify which variable is being assigned, the current domains, and which parent call to Backtrack() it's called within. For example, if the order in which we assign variables is $X_1$, $X_2$, $X_3$, the call stack would be as follows:
      {[01], [01], [01]} $\xrightarrow{X_1=0}$ {0, [01], [01]} $\xrightarrow{X_2=1}$ {0, 1, [01]} $\xrightarrow{X_3=0}$ {0, 1, 0}
      $\xrightarrow{X_1=1}$ {1, [01], [01]} $\xrightarrow{X_2=0}$ {1, 0, [01]} $\xrightarrow{X_3=1}$ {1, 0, 1}

      The notation {1, [01], [01]} means that $X_1$ has been assigned value 1, while $X_2$ and $X_3$ are currently unassigned and each have domain $\{0,1\}$. Note that we omit the comma in the domain for easier reading. We also avoid the weight variable for simplicity; the only possible weights for this problem are 0 and 1. In this case, backtrack is called 7 times. Notice that Backtrack() is not called when there's an inconsistent partial assignment ($\delta=0$); for example, we don't call Backtrack() on $X_2 = 1$ when $X_1$ is already set to 1.

      Using this call stack, we can produce the list of calls in the order they are explored. For this example where we assign variables in order $X_1$, $X_2$, $X_3$, the list would be $\{[01], [01], [01]\}, \{0, [01], [01]\}, \{0, 1, [01]\}, \{0, 1, 0\}, \{1, [01], [01]\}, \{1, 0, [01]\}, \{1, 0, 1\}$.

      Suppose we assign variables in the order $X_3$, $X_1$, $X_2$. Write the list of calls in the order they are explored and draw out the call-stack. How many calls do we make to Backtrack()? Why can this number change depending on the ordering?

    3. We often add heuristics like AC-3 to speed up the backtracking search. How many calls to Backtrack() from your call stack in the previous question would we skip if we use AC-3? Briefly explain why we skip (or don't skip) calls in this search with AC-3.
    For i., a list of all the consistent assignments (1 sentence). For ii., a list of calls in order of exploration, a drawing of the call stack, the number of times Backtrack() is called, and an explanation for why this number can change based on the order in which you assign variables (1-4 sentences). For this problem only you may hand-draw a call stack and paste a picture into the PDF, provided that the drawing is neat and everything is legible. For iii., the number of calls to Backtrack() that get skipped along with an explanation for why we skip these calls with AC-3 (1-2 sentences).
  3. Now let's consider a general case: given a factor graph with $n$ variables $X_1,...,X_n$ and $n-1$ binary factors $t_1,...,t_{n-1}$ where $X_i \in \{0,1\}$ and $t_i(X) = x_i \bigoplus x_{i+1}$. Note that the CSP has a chain structure. Implement create_chain_csp() by creating a generic chain CSP with XOR as factors.

    Note: We've provided you with a CSP implementation in util.py which supports unary and binary factors. For now, you don't need to understand the implementation, but please read the comments and get yourself familiar with the CSP interface. For this problem, you'll need to use CSP.add_variable() and CSP.add_binary_factor().

Problem 1: CSP solving

We'll now pivot towards creating more complicated CSPs, and solving them faster using heuristics. Notice we are already able to solve the CSPs because in submission.py, a basic backtracking search is already implemented. For this problem, we will work with unweighted CSPs that can only have True/False factors; a factor outputs 1 if a constraint is satisfied and 0 otherwise. The backtracking search operates over partial assignments, and specifies whether or not the current assignment satisfies all relevant constraints. When we assign a value to a new variable $X_i$, we check that all constraints that depend only on $X_i$ and the previously assigned variables are satisfied. The function satisfies_constraints() returns whether or not these new factors are satisfied based on the unaryFactors and binaryFactors. When satisfies_constraints() returns False, any full assignment that extends the new partial assignment cannot satisfy all of the constraints, so there is no need to search further with that new partial assignment.

Take a look at BacktrackingSearch.reset_results() to see the other fields which are set as a result of solving the weighted CSP. You should read submission.BacktrackingSearch carefully to make sure that you understand how the backtracking search is working on the CSP.

  1. Let's create an unweighted CSP to solve the n-queens problem: Given an $n\times n$ board, we'd like to place $n$ queens on this board such that no two queens are on the same row, column, or diagonal. Implement create_nqueens_csp() by adding $n$ variables and some number of binary factors. Note that the solver collects some basic statistics on the performance of the algorithm. You should take advantage of these statistics for debugging and analysis. You should get 92 (optimal) assignments for $n=8$ with exactly 2057 operations (number of calls to backtrack()).

    Hint: If you get a larger number of operations or your code times out on the test cases, make sure your CSP is minimal. Try to define the variables such that the size of domain is O(n).

    Note: Please implement the domain of variables as 'list' type in Python. You can refer to create_map_coloring_csp() and create_weighted_csp() in util.py as examples of CSP problem implementations. You can try these examples out by running:

    python run_p1.py

  2. You might notice that our search algorithm explores quite a large number of states even for the $8\times 8$ board. Let's see if we can do better. One heuristic we discussed in class is using most constrained variable (MCV): To choose an unassigned variable, pick the $X_j$ that has the fewest number of values $a$ which are consistent with the current partial assignment ($a$ for which satisfies_constraints() on $X_j=a$ returns True). Implement this heuristic in get_unassigned_variable() under the condition self.mcv = True. It should take you exactly 1361 operations to find all optimal assignments for 8 queens CSP — that's 30% fewer!

    Some useful fields:

Problem 2: Course Scheduling

In this problem, you will leverage our CSP solver for the problem of course scheduling. We have scraped a subset of courses that are offered from Stanford's Bulletin. For each course in this dataset, we have information on which quarters it is offered, the prerequisites (which may not be fully accurate due to ambiguity in the listing), and the range of units allowed. You can take a look at all the courses in courses.json. Please refer to util.Course and util.CourseBulletin for more information.

To specify a desired course plan, you would need to provide a profile which specifies your constraints and preferences for courses. A profile is specified in a text file (see profile*.txt for examples). The profile file has four sections:

Constrained requests. To allow for more flexibility in your preferences, we allow some freedom to customize the requests:

Each request line in your profile is represented in code as an instance of the Request class (see util.py). For example, the request above would have the following fields:

It's important to note that a request does not have to be fulfilled, but if it is, the constraints specified by the various operators after,in must also be satisfied.

You shall not worry about parsing the profiles because we have done all the parsing of the bulletin and profile for you, so all you need to work with is the collection of Request objects in Profile and CourseBulletin to know when courses are offered and the number of units of courses.

Well, that's a lot of information! Let's open a python shell and see them in action:
import util
# load bulletin
bulletin = util.CourseBulletin('courses.json')
# retrieve information of CS221
cs221 = bulletin.courses['CS221']
# look at various properties of the course
print(cs221.prereqs)  # the prerequisites

# load profile from profile_example.txt
profile = util.Profile(bulletin, 'profile_example.txt')
# see what it's about
# iterate over the requests and print out the properties
for request in profile.requests:
    print(request.cids, request.quarters, request.prereqs, request.weight)

Solving the CSP. Your task is to take a profile and bulletin and construct a CSP. We have started you off with code in SchedulingCSPConstructor that constructs the core variables of the CSP as well as some basic constraints. The variables are all pairs of requests and registered quarters (request, quarter), and the value of such a variable is one of the course IDs in that Request or None, which indicates none of the courses should be taken in that quarter. We will add auxiliary variables later. We have also implemented some basic constraints: add_bulletin_constraints(), which enforces that a course can only be taken if it's offered in that quarter (according to the bulletin), and add_norepeating_constraints(), which constrains that no course can be taken more than once.

You should take a look at add_bulletin_constraints() and add_norepeating_constraints() to get a basic understanding how the CSP for scheduling is represented. Nevertheless, we'll highlight some important details to make it easier for you to implement:

  1. Implement the function add_quarter_constraints() in submission.py. This is when your profile specifies which quarter(s) you want your requested courses to be taken in. This is not saying that one of the courses must be taken, but if it is, then it must be taken in any one of the specified quarters. Also note that this constraint will apply to all courses in that request.
  2. Let's now add the unit constraints in add_unit_constraints().
    1. In order for our solution extractor to obtain the number of units, for every course, you must add a variable (courseId, quarter) to the CSP taking on a value equal to the number of units being taken for that course during that quarter. When the course is not taken during that quarter, the unit should be 0.
    2. You must take into account the appropriate binary factor between (request, quarter) and (courseId, quarter) variables.
    3. You must ensure that the sum of units per quarter for your schedule are within the min and max threshold, inclusive. You should use the create_sum_variable() function we've implemented for you; pay careful attention to the arguments.

    Hint: If your code times out, your maxSum passed to create_sum_variable() might be too large.

    Note: Each grader test only tests the function you are asked to implement. To test your CSP with multiple constraints you can add to the profile text file whichever constraints that you want to add and run run_p2.py. Here is an example with profile2b.txt as input:
    python run_p2.py profile2b.txt
    Running this command will print information that may be helpful for debugging, such as profile information, the number of optimal assignments found (along with their weight and the number of times backtrack() is called while solving the CSP), one full optimal assignment, and the resulting course schedule.
  3. Now try to use the course scheduler for any two quarters in the future (or more quarters if you wish, although this might lead to a slower search). Create your own profile.txt (take a look at some of the profile text files included in the assignment's main directory for inspiration) and then run the course scheduler:
    python run_p2.py profile.txt
    If the courses you wish to schedule are not listed in courses.json, feel free to add them in as you please! In addition, feel free to modify course details as well (e.g., you can change the list of quarters that a course is being offered in if it does not match the information on the current year's course calendar). You might want to turn on the appropriate heuristic flags to speed up the computation; in particular, self.ac3 = True applies the arc-consistency heuristic that we implement for you, and you can use your own MCV implementation. Does it produce a reasonable course schedule? Please include your profile.txt and the best schedule in your writeup (you can just paste it into the pdf that you submit); we're curious how it worked out for you! Please include your schedule and the profile in the PDF; otherwise you will not receive credit.
    The profile.txt file (pasted into the pdf with the solutions), the corresponding outputted schedule, and a brief discussion (1-2 sentences) on whether or not it seems reasonable. Please also describe any changes you made to courses.json (if applicable; for example, if you added courses).

Problem 3: Residency Hours Scheduling

  1. Many uses of constraint satisfaction in real-world scenarios involve assignment of resources to entities, like assigning packages to different trucks to optimize delivery. However, when the agents are people, the issue of fair division arises. In this question, you will consider the ethics of what constraints to remove in a CSP when the CSP is unsatisfiable.

    Medical residents are often scheduled to work long shifts with insufficient rest, leading to exhaustion and burnout. This can negatively affect the residents and potentially lead to mistakes that also negatively affect the patients in their care [1]. A hospital could use a constraint-satisfaction approach to try to create a work schedule that respects the “on-call night, day-off, rest period, and total work-hour regulations mandated by the Accreditation Council for Graduate Medical Education, as well as the number of residents needed each hour given the demand (aka number of patients and procedures scheduled) [2]. The constraints are:

    1. One day off every 7 days
    2. Minimum 8 hour rest period between shifts
    3. No more than 80 hours of work per week averaged over a 4 week period
    4. At least 14 hours free of clinical work and education after 24 hours of in-house call
    5. Number of residents needed each hour

    Let’s assume for a given hospital that the constraints listed above were collectively unsatisfiable given the number of residents assigned to that hospital. However, its formulation as an unsatisfiable CSP depends on other factors remaining fixed, such as

    1. The number of total residents
    2. The work to be performed by residents as opposed to other staff
    3. The budget available to hire residents or other staff
    In this case, would you remove one of the numbered constraints 1-5 or advocate that the hospital administration change one of A-C to make the problem solveable? If so, explain which one and give a reason why.

    In 2-4 sentences, you must explicitly state which numbered constraint or lettered factor you would change and justify your choice with a reason that explains why you chose that one.

[1] Akresh-Gonzales et al. Resident Burnout and Well-Being: A Q&A with Educator Benjamin R. Doolittle, MD.

[2] Topaloglu et al. A constraint programming-based solution approach for medical resident scheduling problems. Computers & Operations Research. 2011.