Midterm Flashcards
List 3 of the 6 capabilities needed to pass the Total Turing Test. For the 3 listed capabilities, describe what purpose they serve to pass the Total Turing Test.
Natural language processing - Can understand and communicate using natural language
Computer Vision - Can interpret the environment it’s in
Motor - Can interact with the environment it’s in
What is a rational agent?
An entity that will act for the best outcome by maximizing the performance measure
With the help of a practical example for each type, differentiate between goal agents and utility-based agents
Goal-based: Like a vacuum robot, it will roam until it’s gone over everything possible
Utility-based: Like an autonomous car. It uses information like time and fuel efficiency before choosing the best route
When designing an intelligent agent, we used the PEAS acronym. Apply the PEAS components to the task of creating an Automated Taxi system. Specify two examples for each component.
Performance Measure: Travel Time, Fuel Efficiency
Environment: Time of day for traffic, Weather for Road Conditions
Actuators: Brake/Accel and Steering Wheel
Sensors: Seat occupancy knowing how many passengers, Lidar for processing the road
Differentiate between a fully-observable and a partially-observable environment, giving an example for each.
A fully observable environment would be like a chess game. An agent can see the whole board and all the pieces
A partially observable environment would be the roads for an autonomous car. It can’t see past cars or obstacles. It will have gaps of understanding.
What is a consistent heuristic in informed search algorithms?
It’s a heuristic that estimates the cost without overestimating the true cost.
Describe the two possible admissible heuristics for the 8 tile problem that we discussed in class.
Manhattan: It sums the distance that each tile is from its goal, only by vertical and horizontal moves.
Misplaced Tiles: It counts how many tiles aren’t in the goal state.
Outline the basic steps of the minimax algorithm.
Set up the tree
Evaluate the leaf nodes (terminal nodes)
Backpropagate the utility values
Pick the best values for minimizing and maximizing
What is the key difference between the minimax algorithm and the alpha-beta pruning algorithm?
Minimax explores the whole tree, whereas alpha beta eliminates branches which can’t affect the final decision
In the best case, how much does alpha-beta pruning improve upon minimax?
It can search a tree that’s double the depth within the same time without the pruning.
What do the alpha and beta values represent in alpha-beta pruning?
Alpha is the highest value choice found so far, beta is the lowest.
How does the ordering of nodes affect the efficiency of alpha-beta pruning?
If the better moves are ordered first, it improves the efficiency
Does the alpha-beta pruning method give a more accurate answer in game playing than the minimax algorithm? Explain your reasoning.
No, it only improves the speed of the minimax algorithm.
Using the evaluation function formula, differentiate between A* search and greedy best-first search.
A*: f(n) = g(n) + h(n), Greedy: f(n) = h(n),
where g(n) is the cost from the node n to the root. Greedy will get the answer, but it may not be optimal
