Georgia Institute of Technology
CS 6601 Artificial Intelligence
Final Exam Spring 2025
Duration of Exam: April 24th 2025, 8:00 AM (EDT) – May 1st 2025, 7:59 AM (EDT) Weight: 20% Points: 100 + 5 Extra Credit
Submission: Via Canvas. Unlimited attempts during exam window, only last submission will be used for grading.
Instructions
IMPORTANT*: Please read the instructions in its entirety before beginning the exam. Academic Integrity
During the week-long exam, NO resources outside of this semester’s class materials may be used. It is open- book, open-note, and open-course resources, but it is not open internet (e.g., YouTube, Open Courseware, X-Overflow, Discussions, Forums, LLMs, Search platforms or assistants). We do not allow collaboration with any other intelligent life forms or artificial intelligence (e.g., ChatGPT). Do not discuss or share any materials related to the exam (e.g. this PDF file) anywhere and only acquire the exam through official channels (e.g. Canvas, Edstem, Teaching Staff, GT Github). Sharing on any GT or public/private social media platform (e.g., Ed, Slack) or student communities (e.g. Discord) is a violation of Academic Integrity.
Submitting an attempt to this exam signifies you are aware of and in accordance with the Academic Honor Code of Georgia Tech and the Georgia Tech Code of Conduct. Academic misconduct will not be tolerated and and violations will be submitted to the office of student integrity as outlined in the course syllabus.
Communications
From the teaching staff: As the exam period progresses, the teaching staff may discover issues with the exam or find the need to issue clarifications. All communications will be posted on Edstem in the general clarifications thread and under the specific exam section’s clarification thread. And so, you should check Edstem every day to ensure that no changes / clarifications have been released that would affect your work. There will be a period before the exam deadline after which the teaching staff will not release any further changes / clarifications, this will be announced.
From you: As you complete the exam, you may find that you seek clarification about different exam topics or have general questions. Edstem will be in private mode and you should make a private post on Edstem to ask your questions. It is important that you (1) have checked the clarification threads first to see if your question has been answered before making an Edstem post (2) labeled your post with the correct exam section label (failure to do so may result in the responsible staff missing your post and a delay in response time) (3) group your questions based on subject (e.g. questions related to Search are in one post while questions related to Optimization are in a different post). Do NOT at any point email the teaching staff unless you are explicitly directed by the teaching staff to do so.
The exam is submitted through Canvas. Note that there are two assignments in Canvas, one is for sub- mitting the exam and the other is for optionally uploading any physical work. It is important to recognize that submission of physical work is NOT used for partial credit, it is instead used as evidence that you have not plagiarized in the event that your submission comes under OSI investigation and it is helpful in the event that some technical difficulty causes a loss of your main exam submission.
It is recommended that you download the PDF version of the exam when it is released so that you may have an offline copy in the event you encounter any technical difficulties and for keeping track of your answers that you’d like to input into Canvas as Canvas can be quite finnicky. We also recommend that you periodically submit your answers to Canvas so that you do not accidentally end the exam period with no submission. After you have submitted your final submission and are satisfied, do NOT open the Canvas exam again as that will open a new blank submission and leave you with a score of zero.
Question Formats
You may see the following formats in this exam:
• Numeric Decimal Submission – Never round intermediate steps. Only round your answer at the end if necessary. For example if the questions says to round to 2 decimal places, 3.141 becomes 3.14 while a whole number like 25 can be submitted as 25 (you do not need to submit 25.00).
• Relatively Prime Submission – Several questions in the exam state that the answer can be written in the form of p/q where p and q are relatively prime (coprime) integers and that you should submit the value of p + q. Two integers are considered relatively prime (coprime) if they share no common factors other than 1. For example 14 and 15 are relatively prime (coprime) while 6 and 21 are not because they share a common factor of 3. See the Challenge Questions for examples.
• Single Select – Also known as multiple choice, exactly one answer among the options is correct.
• Multiple Select – At least one answer among the options is correct, select all that apply. We will grade Multiple Select questions with partial credit. For a question with N options worth P points, you will receive P/N points for each answer choice that is correctly checked / unchecked. Note that Canvas does not have an automated option for this and we will run a regrading script after the exam is over to enact this.
• Matching – There will be an option bank that you can drag to match with answer choices. Note that an option may be used more than once.
Grading and the Challenge Period
Your last attempt will constitute your graded exam. Do not have an unfinished exam active when the deadline comes, as the system will auto-submit your current working exam. This means that if you had 80 points of completed work on your previous submission, you begin a new attempt, and are only 25% through the exam when the deadline comes, your exam grade may only total 25 points. You will not be able to see the points earned in each submission until after the exam has closed. Always assume that there will be no partial credit.
It is your responsibility to ensure that you have submitted the correct answers to Canvas. Again, we will NOT be looking at your physically submitted work for grading purposes other than plagiarism. No partial credit is provided for issues like accidentally selecting the wrong answer, rounding error, and typos. Exten- sions are also not provided for any reason unless you have reached out with approved ODS accommodations. There have been exceptions in the event of major events (e.g. natural disasters) affecting multiple students, so please reach out to the teaching staff via Edstem if you have a developing situation. Do NOT wait until the last minute to contact the teaching staff.
After the exam period is over, there will be a brief Challenge Period during which the teaching staff will release the answer key and corresponding solutions and students will have the opportunity to challenge the accepted solutions. It is an opportunity for students to outline why they think the accepted solution is incorrect or why another solution should also receive credit based on an alternate interpretation of the question. There is no guarantee that changes will be made, final decision is up to teaching staff discretion. Grades seen in Canvas will not be accurate during this period, you should use the answer key to calculate your expected grade so that you can check that you have the correct grade after the Challenge Period ends and the final regrading script has been run.
Good Luck!
We wish you the best of luck on the exam! This is a great opportunity for you to revisit topics covered in the course material and to consolidate your understanding. And as always, do not hesitate to reach out on Edstem whenever you have questions.
Acknowledgement [1pt]
Please enter your full name below as acknowledgement that you have fully read and understood the exam instructions above.
Respond with your full name.
1 Search [7pts] 6
1.1 SearchQ1[1pt]…………………………………….. 7 1.2 SearchQ2[1pt]…………………………………….. 7 1.3 SearchQ3[1pt]…………………………………….. 7 1.4 SearchQ4[1pt]…………………………………….. 8 1.5 SearchQ5[1pt]…………………………………….. 8 1.6 SearchQ6[1pt]…………………………………….. 8 1.7 SearchQ7[1pt]…………………………………….. 9
2 Game Playing [6pts] 10
2.1 GamePlayingQ1[1pt]…………………………………. 11 2.2 GamePlayingQ2[1pt]…………………………………. 12 2.3 GamePlayingQ3[1pt]…………………………………. 12 2.4 GamePlayingQ4[1pt]…………………………………. 12 2.5 GamePlayingQ5[2pts] ………………………………… 13
3 Optimization [7pts] 14
3.1 OptimizationQ1[1pt] …………………………………. 15 3.2 OptimizationQ2[2pts]…………………………………. 15 3.3 OptimizationQ3[2pts]…………………………………. 15 3.4 OptimizationQ4[2pts]…………………………………. 15
4 Constraint Satisfaction Problems (CSPs) [7pts] 16
4.1 CSPQ1[1pt]……………………………………… 18 4.2 CSPQ2[1pt]……………………………………… 18 4.3 CSPQ3[1pt]……………………………………… 18 4.4 CSPQ4[1pt]……………………………………… 19 4.5 CSPQ5[1pt]……………………………………… 19 4.6 CSPQ6[1pt]……………………………………… 19 4.7 CSPQ7[1pt]……………………………………… 19
5 Probability [6pts] 20
5.1 ProbabilityQ1[1pt] ………………………………….. 20 5.2 ProbabilityQ2[1pt] ………………………………….. 20 5.3 ProbabilityQ3[1pt] ………………………………….. 21 5.4 ProbabilityQ4[1pt] ………………………………….. 21 5.5 ProbabilityQ5[1pt] ………………………………….. 22 5.6 ProbabilityQ6[1pt] ………………………………….. 22
6 Bayes Nets [7pts] 23
6.1 BayesNetsQ1[1pt] ………………………………….. 24 6.2 BayesNetsQ2[1pt] ………………………………….. 24 6.3 BayesNetsQ3[1pt] ………………………………….. 25 6.4 BayesNetsQ4[1pt] ………………………………….. 26 6.5 BayesNetsQ5[1pt] ………………………………….. 26 6.6 BayesNetsQ6[1pt] ………………………………….. 27 6.7 BayesNetsQ7[1pt] ………………………………….. 27
7 Machine Learning [15pts] 28
7.1 MachineLearningQ1[2pts]………………………………. 28 7.2 MachineLearningQ2[2pts]………………………………. 28 7.3 MachineLearningQ3[2pts]………………………………. 28 7.4 MachineLearningQ4[2pts]………………………………. 29 7.5 MachineLearningQ5[2pts]………………………………. 29 7.6 MachineLearningQ6[2pts]………………………………. 29 7.7 MachineLearningQ7[3pts]………………………………. 30
8 Pattern Recognition Through Time [15pts] 31
8.1 PatternRecognitionThroughTimeQ1[3pts] ……………………… 33 8.2 PatternRecognitionThroughTimeQ2[3pts] ……………………… 33 8.3 PatternRecognitionThroughTimeQ3[3pts] ……………………… 34 8.4 PatternRecognitionThroughTimeQ4[3pts] ……………………… 34 8.5 PatternRecognitionThroughTimeQ5[3pts] ……………………… 35
9 Deep Learning [15 pts] 36
9.1 DeepLearningQ1[3pts]………………………………… 37 9.2 DeepLearningQ2[3pts]………………………………… 37 9.3 DeepLearningQ3[3pts]………………………………… 37 9.4 DeepLearningQ4[3pts]………………………………… 38 9.5 DeepLearningQ5[3pts]………………………………… 38
10 Planning Under Uncertainty [14pts] 39
10.1PlanningUnderUncertaintyQ1[2pts] …………………………. 39 10.2PlanningUnderUncertaintyQ2[2pts] …………………………. 39 10.3PlanningUnderUncertaintyQ3[2pts] …………………………. 40 10.4PlanningUnderUncertaintyQ4[2pts] …………………………. 40 10.5PlanningUnderUncertaintyQ5[2pts] …………………………. 40 10.6PlanningUnderUncertaintyQ6[2pts] …………………………. 41 10.7PlanningUnderUncertaintyQ7[2pts] …………………………. 41
11 Deep Learning Extra Credit [5pts] 42
11.1DeepLearningExtraCreditQ1[1pt]………………………….. 43 11.2DeepLearningExtraCreditQ2[1pt]………………………….. 43 11.3DeepLearningExtraCreditQ3[1pt]………………………….. 43 11.4DeepLearningExtraCreditQ4[1pt]………………………….. 43 11.5DeepLearningExtraCreditQ5[1pt]………………………….. 44
程序代写 CS代考 加QQ: 749389476
1 Search [7pts] Introduction
Image scaling is an important task for image processing. An intuitive approach to do image scaling is applying a scaling factor directly to the images, and reassigning the pixel values according to the scaling factor. However, this can lead to distortion to the objects in the images. To address this issue, seam carving is a content-aware image scaling technique in computer vision. Figure 1 shows the power of seam carving. The leftmost image is the original image, and the middle image illustrates how the castle gets distorted by simply rescaling the image. The rightmost image shows the results achieved by seam carving.
Figure 1: Comparison of Resized Images and Seam Carved Images
How does the magic happen? An image is represented as a 2D grid where each pixel has an associated energy value. The energy values are typically defined by the summation of absolute values of image gradients, i.e.,
Energy = | ∂ Image| + | ∂ Image| ∂x ∂y
A seam is a connected path of pixels (typically vertical: from top to bottom; or horizontal: from left to right) such that each pixel in the seam is adjacent (or diagonally adjacent) to the next. The goal is to find a seam that minimizes the cumulative energy, thereby preserving important image features when rescaling. This problem can be solved by dynamic programming, while we are going to solve this by using search algorithms in this question.
Figure 2: Sample grid of pixels and their energy values
The example in Figure 2 illustrates a 5 (rows) by 6 (cols) image (a very small image), and the numbers in the cells are the energy values (yes, you don’t need to calculate the energy value by yourself). The goal
of seam carving is to find a path (seam) from top to bottom or from left to right where the pixels of the seam should be either adjacent or diagonally adjacent to the next pixel. For example, if we are trying to find a path from top to bottom, the connected pixels of A3 are B2, B3, and B4. And there will be only two connected pixels for an edge pixel, which means the connected pixels of B6 are C5 and C6. The path (seam) is considered optimal if the energy is minimal.
1.1 Search Q1 [1pt]
When seam carving, which of the following algorithms are guaranteed to find the optimal solution (seam)?
Multiple select.
(a) DFS (b) BFS (c) UCS
1.2 Search Q2 [1pt]
If we start from pixel A4 in Figure 2, and want to find a seam (path) with minimum energy values from top to bottom, we can consider this as a minimum path finding problem. To solve this problem, the pixels in the image can be considered as nodes, and the edges are connected between the pixels according to the rules defined in the introduction above. The weight of the edge is given by the energy value of the target node of the edge (e.g., the weight of the edge between A2 and B2 is 9, which is the energy value of B2). In addition, we can further introduce a target node t, where node t is connected to nodes E1, E2, E3, E4, E5, E6 with edge weights of zero.
Which of the following nodes are in the seam with minimum energy starting from pixel A4 in Figure 2?
Select one cell per row (5).
1.3 Search Q3 [1pt]
Given the information in the introduction, we want to design a heuristic so that we can use the A* algorithm to solve this problem. Assuming our start and goal are nodes A4 and t respectively (as introduced in Search Q2), we develop a non-negative heuristic function h1(n) where h1(t) = 0 and:
• h1(A4)=5 • h1(B3)=4 • h1(B4)=5 • h1(B5)=6
Which of the following statements about h1(n) are true?
Multiple select.
(a) h1(n) might be admissible (b) h1(n) is NOT admissible
(c) h1(n) might be consistent (d) h1(n) is NOT consistent
1.4 Search Q4 [1pt]
Given the information in the introduction, we want to design a heuristic so that we can use the A* algorithm to solve this problem. Assuming our start and goal are nodes A4 and t respectively (as introduced in Search Q2), we develop a non-negative heuristic function h2(n) where h2(t) = 0 and:
• h2(A4)=8 • h2(B3)=6 • h2(B4)=7 • h2(B5)=4
Which of the following statements about h2(n) are true?
Multiple select.
(a) h2(n) might be admissible (b) h2(n) is NOT admissible
(c) h2(n) might be consistent (d) h2(n) is NOT consistent
1.5 Search Q5 [1pt]
Given the information in the introduction, we want to design a heuristic so that we can use the A* algorithm to solve this problem. Assuming our start and goal are nodes A4 and t respectively (as introduced in Search Q2), we develop a non-negative heuristic function h3(n) where h3(t) = 0 and:
• h3(A4)=7 • h3(B3)=3 • h3(B4)=5 • h3(B5)=5
Which of the following statements about h3(n) are true?
Multiple select.
(a) h3(n) might be admissible (b) h3(n) is NOT admissible
(c) h3(n) might be consistent (d) h3(n) is NOT consistent
1.6 Search Q6 [1pt]
Now, consider a more general case, we want to find a seam with minimum energy value from top (any pixel) to bottom (any pixel), we will introduce two nodes, s and t. Node s is the start node that connects to A1, A2, …, A6 and t is the end node that connects to E1, E2, …, E6. How many edges are there in this newly built graph (including the edges involving s or t)?
Answer as a whole number.
1.7 Search Q7 [1pt]
Let us apply the same trick to find a seam from left to right (instead of top to bottom), where we will again we introduce two nodes, s and t. Node s is the start node that connects to A1, B1, …, E1 and t is the end node that connects to A6, B6, . . . , E6. How many edges are there in this newly built graph (including the edges involving s or t)?
Answer as a whole number.
CS Help, Email: tutorcs@163.com
2 Game Playing [6pts] Part I
“Multiplayer Online Battle Arena” (MOBA) games have become an immensely popular gaming genre, characterized by their strategic, team-based gameplay on dynamic maps. In a typical MOBA, two opposing teams compete to achieve objectives such as destroying the enemy base, eliminating opponents, or securing key resources. This type of game typically requires making optimal decisions in an uncertain and dynamic environment as players need to predict opponent’s actions and adapt to random events. In this problem, we will explore a simplified two-player MOBA scenario and apply AI planning techniques to model and optimize decision-making in this complex, probabilistic environment.
Figure 3: The MOBA game world for this problem
Consider a 5×4 2D grid map, shown in Figure 3 above, where P1 and P2 represent two players from opposing teams. The black cells on the grid are obstacles that players cannot move onto. The objective of each player is to eliminate the opponent.
Each player has the following attributes:
• Health: Both players start with 50 health points (HP) A player dies when their HP falls to 0. There’s no “overkill”, so a player’s HP will never be negative.
• Actions: They can perform one of the following potential moves when it is their turn:
– Movement: Players can move in one of 8 directions (up, down, left, right, and diagonals: up-left, up-right, down-left, and down-right). They can occupy the same cell. If a movement will take the player outside the grid or onto obstacle cells, that movement is not an option for the player.
– Basic Attack: Players can attack with a simple skill that deals small damage but has a high accuracy rate.
– Trick Shot: Alternatively, they can use a trick shot, which deals greater damage but has a lower hit rate. Regardless of whether the attack hits or misses, using this skill reduces the player’s own health by a small amount. The player can die from casting this skill.
The industry methods for AI players in MOBA games typically involve finite state machines, heuristic planning, or in some cases, deep reinforcement learning and imitation learning. In the case of this game, we will use Expectiminimax game trees to see how an AI agent would work in a simplified MOBA setting. We will use the following specifications to build the game trees:
• Consider that regular triangles denote the max player, while inverse triangles denote the min player. The circular nodes denote chance nodes, and the number on a line means the probability of entering that branch.
• The evaluation function will be max player HP – min player HP.
• Consider that the agent will not “do nothing”. This means it will either move in a valid direction or
attack the opponent using one of the two attack methods.
• For this problem, you will always be playing from the perspective of P1 (the max player).
2.1 Game Playing Q1 [1pt]
An agent with no knowledge of the map will try all possible moves at each state. However, we can hardcode the map into the agent, so that it will not walk off the board or walk into the obstacle. If the map is hardcoded into the agent, what is the average branching factor for this agent?
Select one.
(a) 13/3 (b) 40/9
(c) 6 (d) 58/9
Suppose the following tree depicts a particular state of the game, and the agent only studies the potential outcomes of performing basic attack or trick shot.
Figure 4: Game Tree for Game Playing Q2 – Q4
2.2 Game Playing Q2 [1pt]
What is the value of node E?
Answer as an exact decimal. Do NOT round.
2.3 Game Playing Q3 [1pt]
What is the value of node C?
Answer as an exact decimal. Do NOT round.
2.4 Game Playing Q4 [1pt]
What is the value of node A?
Answer as an exact decimal. Do NOT round.
程序代写 CS代考 加微信: cstutorcs
2.5 Game Playing Q5 [2pts]
For simplicity, we further reduce the actions as discussed in Q1 so that there are only 4 general actions: • Basic attack
• Trick shot
• Move towards the opponent
• Move away from the opponent
Figure 5: Game tree for Q5
This tree in Figure 5 depicts the game at a certain stage, with the newly defined four general actions. In this game tree, how many leaf nodes will be pruned, using alpha-beta pruning from left to right? Note that this is an expecti-minimax game tree.
Answer as a whole number.
3 Optimization [7pts] Introduction
Genetic Algorithms can be used to find the maximum of a function. For this problem, the function we want to maximize is a function of two variables f(x,y) = |0.2x∗sin(x+y)| + |(x+y)∗cos(y2)/10| in the interval x, y ∈ [0, 1023]. The input can only take integer values.
The Algorithm
To find the minimum of this function, we will do the following steps.
1. Create an initial population of 10. This population stays the same throughout successive generations. The population consists of inputs for the function we are trying to optimize. The inputs are selected at random from the domain x ∈ [0, 1023] and y ∈ [0, 1023]. This population will stay the same throughout each generation.
2. Calculate the fitness of the population. The fitness of (x, y) is just the value of f at (x, y).
3. Calculate the probability of selection. Selection is proportional to fitness, i.e. if one member has a fitness twice as high as another member, it is two times more likely to be selected. Five pairs are chosen. The same parent can appear in multiple pairs.
4. Inputs are encoded as 10 bits. For example, x = 4 is encoded as x = 0000000100 and y = 1010 is encoded as y = 0001100100. This will allow us to do crossover.
5. A one-point crossover is performed for x and y of each parent, producing two children with 50 percent of their bits inherited from each parent. For example, if Parent 1 = (40, 90) and Parent 2 = (150, 400) then Child 1 = (54, 80) and Child 2 = (136, 410).
6. Mutation is performed by choosing a bit at random with probability = p and flipping that bit. For example, a possible mutation could be x = 0000000100 mutated into x = 0000000110 where the second to last bit is flipped from 0 to 1. An individual may have more than one bit flipped during mutation.
7. After crossover and mutation, a new generation is created. This process is repeated for 50 generations. After 50 generations, the input with the minimum fitness is selected.
3.1 Optimization Q1 [1pt]
What is the fitness of f(100,40)?
Answer as a decimal rounded to 6 decimal places.
3.2 Optimization Q2 [2pts]
From an initial population of 10 we have these inputs and fitness values shown in Table 1. What is the probability of selecting the input (3, 5) for the crossover?
57 87 5 13
29.176346 7.362679 1.386577 61.256061
535 61 99 106 10 80
400 61 91 639
96.791633 16.076571 9.336979 68.072526 41.567432
600 206.558638
Table 1: Input Variables (x, y) and Corresponding Fitness Values Answer as a decimal rounded to 6 decimal places.
3.3 Optimization Q3 [2pts]
The inputs (456, 18) and (903, 61) are selected as parents of the next generation. What are the two children of the parents as outlined in the Algorithm?
Select one.
(a) (51, 89) and (750, 25) (b) (679, 62) and (270, 21) (c) (455, 29) and (904, 50)
(d) (305, 180) and (689, 30)
3.4 Optimization Q4 [2pts]
If a target of an average of at least 1 mutation per population per generation is desirable, what should the mutation probability be?
Answer as a decimal rounded to 6 decimal places.
4 Constraint Satisfaction Problems (CSPs) [7pts] Introduction
Determining the order of the nodes and values is a critical direction of research in solving of Constraint Satisfaction Problems (CSPs). As you will see in this problem, the ordering of nodes and values can have a strong impact on the solve time of the CSP. We will be using map coloring for simplicity, but you can imagine the differences you are seeing here as you increase the number of nodes and edges in more general problems. While we will be sticking with heuristics in this question, you can check out a related paper Learning Variable Ordering Heuristics For Solving Constraints Satisfaction Problems that uses graph neural networks (GNNs) to improve on node and value ordering heuristics. This paper is the motivation for this question but will not assist with you solving it.
Figure 7: Map Coloring Graph
The graph for all problems is given in Figure 7. We will be coloring this map using three different colors,
denoted by {1, 2, 3} for clarity.
Unary Constraints: • C ̸= 1
Binary Constraints: If two nodes have an edge connecting them in Figure 7 they cannot be given the
same assignment (color). As an example if A = 1 then F ̸= 1 and D ̸= 1.
Unless otherwise instructed, you should assume that variables are chosen in alphabetical order and colors in numerical order ascending. We will be trying out different heuristics for node ordering and value ordering to see if we can improve the solve time of the CSP.
We define the number of steps the algorithm has to take as the number of times we make an assignment of a value to a variable, where the assignment results in a consistent (does not violate constraints) partial assignment or consistent complete assignment.
As an example, we provide the following solution to backtracking without any forward checking using a smaller subgraph.
Figure 8: Example Map Coloring Graph – Do not use for questions in this section
Unary Constraints: • I ̸= 1
Binary Constraints: If two nodes have an edge connecting them in Figure 8 they cannot be given the same assignment (color). As an example if H = 1 then J ̸= 1 and K ̸= 1.
Example Solution: ⇒ 9 steps of backtracking without any forward checking to find the solution ∅
Figure 9: Visualization of the tree generated by the backtracking search algorithm performed on the graph in Figure 8 with all consistent partial assignments shown with counts. S1 is step 1, S2 is step 2, etc. The solution takes 9 steps to find. The leafs do not count since there is no consistent assignment available for K denoted by the empty set symbol ∅.
4.1 CSP Q1 [1pt]
Using the Backtracking algorithm (AIMA 4th Ed. Section 5.3 Figure 5), how many steps (as defined in the instructions) does it take to find a complete and consistent assignment on the map coloring graph seen in Figure 7? Perform the algorithm without any type of forward checking but with unary constraints applied, nodes selected in alphabetical order for assignment, and colors selected in ascending numerical order for value assignment.
Select one.
(a) 7 (b) 10 (c) 15 (d) 18
4.2 CSP Q2 [1pt]
How many values are removed by applying Forward Checking (FC) (AIMA 4th Ed. Section 5.3.2 ) on the graph when starting with D = 3 but with no other variables assigned values on the graph in Figure 7? The initial graph should already have unary constraints applied (i.e. unary constraints do not count as removing a value or towards your reported count).
Select one.
(a) 1 (b) 3 (c) 6 (d) 9
4.3 CSP Q3 [1pt]
Using the Backtracking algorithm (AIMA 4th Ed. Section 5.3 Figure 5), how many steps (as defined in the instructions) does it take to find a complete and consistent assignment on the map coloring graph seen in Figure 7? Perform the algorithm with forward checking and with unary constraints applied, nodes selected in alphabetical order for assignment, and colors selected in ascending numerical ord