Actionability in AI Flashcards
What is the role of search algorithms in AI?
Search algorithms help AI take actions in real-world situations by providing solutions through a sequence of actions transforming the start state to the goal state.
How do deep learning and search algorithms complement each other in robotics?
Deep learning helps robots understand their surroundings, while search algorithms help plan the best route to solve a problem.
What are the key components of a search problem?
Initial state, State space, Actions, Goal state, Goal test, Path cost.
In a search space, what do the root, branches, and nodes represent?
Root: initial state, Branches: actions, Nodes: outcomes of actions
What are the four main properties of a search algorithm?
Completeness, Optimality, Time Complexity, Space Complexity.
What are the two main types of search algorithms?
Uninformed (blind) search and Informed search.
Name three uninformed search algorithms.
Breadth-first search, Depth-first search, Uniform cost search.
How does Breadth-First Search (BFS) work?
It explores all nodes at the current depth before moving to the next depth level.
Is BFS complete and optimal?
Yes, if all actions have the same cost.
How does Depth-First Search (DFS) work?
It explores as far as possible along each branch before backtracking.
Is DFS complete and optimal?
No, DFS may not find a solution (infinite loops) and doesn’t guarantee the shortest path.
What is the strategy of Uniform Cost Search (UCS)?
It expands the least cost node first, where cost is the cumulative sum from the root.
Is UCS complete and optimal?
Yes, it finds the lowest-cost solution.
What distinguishes informed search from uninformed search?
Informed search uses heuristics to guide the search.
How does Greedy Search work?
It selects the node closest to the goal based on heuristic h(x), ignoring path cost.
Is Greedy Search complete or optimal?
No, it may miss the solution and is not guaranteed to be optimal.
What is the formula used in A* Search?
f(x) = g(x) + h(x), where g(x) is the path cost so far, and h(x) is the estimated cost to the goal.
Is A* Search complete and optimal?
Yes, it is both complete and optimal with an admissible heuristic.
What is the 8-puzzle problem in AI?
A puzzle with 9 tiles (one empty) arranged in a 3x3 grid; the goal is to reach a specific tile configuration.
What is the Dropout Monte Carlo method used for in AI?
It quantifies uncertainty by running multiple stochastic forward passes and analyzing output variability.
What are the steps in Dropout Monte Carlo for uncertainty estimation?
Train with dropout, run multiple inferences, average predictions, and compute standard deviation.
What is XAI (Explainable AI)?
Techniques to interpret and explain how AI models make decisions, especially black-box models.
What is Grad-CAM?
A technique that highlights important regions in an image that influenced the prediction of a CNN.
What are the steps of Grad-CAM?
Compute class gradients w.r.t. last conv layer, take weighted sum of feature maps, upscale to image size, overlay heatmap.
