16312467
Beschreibung
Karteikarten von Thomas Scott, aktualisiert more than 1 year ago
|
Erstellt von Thomas Scott
vor fast 6 Jahre
|
|
| Frage | Antworten |
| Agent | Anything that can be viewed as perceiving its environment through sensors and acting upon that environment through actuators |
| Agent Program | Physical architecture that instantiates an agent function |
| Rational Agent | Maximizes the expected value of a performance measure given the observed percept sequence |
| Mode-based Agent | Maintains an internal representation of the world for reasoning in partially observable environments |
| Utility-based Agent | Selects appropriately between several goals based on likelihood of success |
| Goal-based Agent | Reasons about future actions in order to reach a desirable outcome |
| Agent Function | Maps percept histories into actions |
| Random Actions | Useful for exploring unknown environments |
| Turing Test | Requires natural language processing and automated reasoning |
| During Classical search, nodes which are queued for expansion are part of the _______ region of the search space, while nodes which have been expanded are part of the _______ region | 1) frontier 2) explored |
| True or False: In classical search methods, the search strategy determines the order in which nodes are expanded during the search process. | True |
| In complexity theory, _________ problems are a subset of problems in NP to which any problem in NP can be reduced in __________ time. | 1) NP complete 2) P |
| The PEAS acronym stands for ________, _________, ___________, and _________ | 1) Performance 2) Environment 3) Actuators 4) Sensors |
| True or False: Any solution found by greedy best-first search must be optimal | False |
| Explain the difference between uniformed search strategies and informed search strategies | Uninformed strategies do not care how the goal is achieved, while informed strategies take heuristic estimates into consideration |
| What are the "four corners" of AI | Thinking Humanly Acting Humanly Thinking Rationally Acting Rationally |
| What does it mean for a heuristic to be admissible? | It has to be less than the actual cost and is optimistic |
| What does it mean for a heuristic to be consistent(monotonic)? | f always increases along any path to the goal |
| What does it mean for a heuristic to be dominant over an alternative heuristic? | h2(n) >= h1(n) for all n and both are admissible |
| Describe a search space in which Iterative Deepening Search performs better than Depth-First search | If there is an infinte deapth IDS would be better. |
| Describe when and why Iterative Deepening Search would be preferred over Breadth First Search for some search problems. | When space is limited, IDS is the better option |
| Explain when and why using a closed list for A* Search is not preferred for some search problems | If h(n) is inconsistent it would lead to sub optimal solution |
| Describe a problem environment that is partially observable. | The game battleship |
| Describe a problem where the environment is fully observable, but the agent is unable to perform optimally | In a dynamic environment, even if the environment is fully observable, the agent can not perform optimally |
| Suppose there is an environment in which a rational agent selects actions from a learned probability distribution (based on its past experience with the environment). Is this an example of a problem with stochastic environments? | [Fill in later] |
| Suppose a learning agent is asked to solve a task multiple times, starting over from the same initial state each time. Is this an example of an episodic environment or sequential environment? | Episodic |
| How could you transform an episodic environment into a sequential environment? | You could add an element of time to make the environment sequential. (From slot machine example) Instead of the environment resetting every time the handle is pulled, the environment would consist of many pulls |
| Multi-agent environments give rise to (at least) two kinds of interactions among the agents that are not found in single-agent environments. Name two interactions and describe why they are advantageous/disadvantageous from an agent's perspective | 1) Adversarial, agents must act against one another. This can be used to help agents learn. Example: Two agents learning chess by playing one another 2) Cooperative, agents must work together to complete a task. This can help up the effectiveness of completing a task |
| Use the four steps of problem formulation to explicitly define the problem faced by a vacuum-cleaner agent in an environment with exactly two locations. | 1) Check the state 2) Check the actions 3) Check the goal 4) Check step costs |
| What are some common problems with greedy local search? | It can get stuck on plateaus and local maxima |
| How does simulated annealing attempt to solve the common problems with greedy local search? | It allows for bad moves |
| How is the temperature parameter in simulated annealing used to aid exploration? | [Fill In] |
| What are three mechanisms used by genetic algorithms to overcome the problems of greedy local search? | 1) mutation 2) selection 3) cross over |
| What is the fundamental difference between the local search approaches above, and the classical search methods (DFS, A*, etc) that we studied earlier in the semester? | Only concerned with finding a goal/solution in local search instead of a path which is the aim of classical search |
| What is the horizon effect in adversarial search? | At some point a search has a max depth, and cannot see past a horizon |
| What happens when we are playing a game using the minimax algorithm and the opponent (MIN) makes a non- optimal move? | This lets the player(MAX) makea better move |
| Give an example of a valid sentence in propositional logic | p v -p |
| Give an example of an unsatisfiable sentence in propositional logic | p <=> -p |
| Is the sentence (-(-Smoke => -Fire) => (Fire => Smoke)) valid, unsatisfiable , or satisfiable ? | satisfiable |
| Linear regression is an example of parametric or nonparametric model? | parametric |
| True or False: When constructing a non-parametric model, the required number of basis vectors or functions can be determined directly from the raining data | False |
| True or False: Non- parametric models may be constructed using an arbitrary number of parametric models. | True |
Möchten Sie mit GoConqr kostenlos Ihre eigenen Karteikarten erstellen? Mehr erfahren.