Project 3 Grading Details Question 1: The auto-grader will run on the maze smallHunt with a random ghost. The auto-grader will be comparing the belief distribution generated by your agent with one produced by our solution agent. If the difference between the two distributions is within some floating point threshold (squared norm, ~0.01), it will be considered correct. Multiple such comparisons will be made with different states and observations. Partial credit will be given for < 10 errors, while full credit will be given for 0 errors. Question 2: Basically the same as Question 1 except the auto-grader will be checking with elapse on, meaning that the belief distribution of the agent will be updated with elapseTime. Same margin of error and partial credit policy. Question 3: The auto-grader will run 10 games on smallHunt with multiple random ghosts. Your agent will be expected to score above 700 on at least 8 of these for full credit. For partial credit, you'll need to score above 700 on at least 4. Question 4: The auto-grader will be running the solution greedy buster agent with your particle inference module on oneHunt with a random ghost for 10 games. You will need to score at least 100 points each game. Grade points will be assigned as follows (count = number of games out of 10 you score >= 100 in): count >= 8: full credit count >= 6: 3 points count >= 4: 1 point count < 4: 0 points Question 5: The auto-grader will be running DispersingGhost ghosts on teenyHunt with 1,800 particles and will compare your MarginalInference module with the solution MarginalInference module, which should have elapseTime implemented. If the difference between the two generated belief distributions is within some floating point threshold (squared norm, ~0.01), it will be considered correct. Multiple such comparisons will be made. Partial credit will be given for < 10 errors, while full credit will be given for 0 errors. Question 6: The auto-grader will be testing for both correctness (3 pts) and functionality (3 pts) on this question. For correctness, the auto-grader will be running your particle filter inference module on teenyHunt with 1,800 particles and DispersingGhost ghosts, comparing the generated belief distributions with those generated by the solution. If the difference between the two generated belief distributions is within some floating point threshold (squared norm, ~0.2), it will be considered correct. Multiple such comparisons will be made. Partial credit will be given for < 10 errors, while full credit will be given for 0 errors. For functionality, the auto-grader will be running your particle filter inference module with the solution greedy busters agent on oneHunt with DispersingGhost ghosts. After 10 games, the average score will be calculated. If your average score > 480, you will receive full credit. No partial credit will be given for this part of the question.