Goldsmiths: Artificial Intelligence

27 flashcards · Computer Science · by lakesare

1. What's an open-loop control system?

   Systems that don't have any feedback loops. These type of systems doesn’t depend upon its output i.e. in open loop feedback systems, output is not used as the control variable for the system. Open loop systems are one way signal flow systems.As these systems doesn’t contain a feedback loops ,these are also known as Non-feedback systems. As there is no feedback loop, no data is transferred back to the open loop systems.

2. Name 2 examples of search problems. An important aspect of intelligence is goal-based problem solving. Several problems can be formulated as finding a sequence of actions that lead to a desirable goal.Each action changes the state and the aim is to find the sequence of actions and states that lead from the initial state to a final (goal) state.

   vacuum world The traveling salesperson problem (TSP) is a touring problem in which each city TRAVELING SALESPERSON PROBLEM must be visited exactly once. The aim is to find the shortest tour. The problem is known to be NP-hard, but an enormous amount of effort has been expended to improve the capabilities of TSP algorithms. In addition to planning trips for traveling salespersons, these algorithms have been used for tasks such as planning movements of automatic circuit-board drills and of stocki…

3. [search problems performance] We can evaluate an algorithm's performance in four ways: completeness optimality time complexity - how long does it take to find a solution? space complexity - how much memory is needed to perform the search? What's completeness VS optimality?

   completeness - if at least one solution exists then the algorithm is guaranteed find a solution in a finite amount of time. optimality - when algorithm finds a solution it finds the best solution.

4. What are the advantages of breadth-first search (BFS) over depth-first search (DFS)?

   BFS is complete and optimal, while DFS is not guaranteed to halt when there are loops

5. What is the advantage of DFS over BFS?

   If m is the maximum path length and b is the branching factor, the space complexity for DFS is mb while for BFS it is bm.

6. What's uninformed VS informed search? Uninformed search, aka blind search.Informed search, aka heuristic search.

   Informed search knows what non-goal states are more promising.

7. Name states in a well-defined search problem.

   Initial state; Operator or successor function - the set of states reachable from any state x with one action; State space - all states reachable from initial state by any sequence of actions; Path - sequence through state space; Path cost - function that assigns a cost to a path. Cost of a path is the sum of costs of individual actions along the path; Goal test - test to determine if at goal state.

8. In the 8-puzzle game, describe states operator function goal path cost 8-puzzle game

   States: location of blank and location of the 8 tiles Operator function: blank moves left, right, up and down Goal: match the state given by the Goal state Path Cost: each step has the cost 1; total cost is considered as being the length of path.

9. What's the branching factor?

   Number of children at each node. In computing, tree data structures, and game theory, the branching factor is the number of children at each node, the outdegree. If this value is not uniform, an average branching factor can be calculated.

10. Build the search-tree for the 8-puzzle game. 8-puzzle game

11. What's the difference between classification & regression? Regression and classification are categorized under the same umbrella of supervised machine learning. Both share the same concept of utilizing known datasets (referred to as training datasets) to make predictions.

    Classification - output is categorical (or discrete). classifier.predict([[120, 1]])# Output is 0 for appleRegression - output is numerical (or continuous). np.sqrt(metrics.mean_squared_error(y_test, y_predict))# Output gives the RMSE number as 1.4046514230328955

12. What's overfitting?

    When decision tree is perfectly fitted to the given data, but generalises badly to new data.

13. What does MDP stand for?

    Markov Decision Process

14. How does Markov Decision Process relate to Markov chains?

    MDP is an extension of Markov chains; the difference is the addition of actions (allowing choice) and rewards (giving motivation). Conversely, if only one action exists for each state (e.g. "wait") and all rewards are the same (e.g. "zero"), a Markov decision process reduces to a Markov chain.

15. What's Markov property?

    For any state s , all the information the agent needs to make a decision is contained in s.

16. Markov Decision Process is described by a tuple: M = <S, A, P, C>. What does each letter stand for?

    S - set of all states agent may find itself in A - set of all actions that the agent may execute P(s' | s, a) - transition function => probability of moving to state s', given we have state s and execute an action a. C(s, a) - reward function => scalar reward for for executing action a when we have state s.

17. What does SARSA stand for?

    State–action–reward–state–action SARSA is an algorithm for learning a Markov Decision Process policy.

18. Give the example of contradiction in terms.

    True fiction. A statement or group of words associating objects or ideas that are incompatible.

19. Describe applications of this ANN: perceptron multilayered feed-forward network Kohonen-Grossberg counterpropagation Boltzmann machines Hopfield nets

    Perceptron can be used to classify the data (of any dimensions) into two parts. Limitations: The neuron is a binary classifier since it can only learn to distinguish between two classes (e.g. blue and red) max. The neuron is a linear classifier because it’s decision boundary is a straight line for 2D data (or a flat plane for 3D data, etc).

20. [Logic] What's a model?

    A model is a particular interpretation of variables. Each model has a yellow background.

21. [Logic] When is a set of sentences consistent with the model? Each model has a yellow background.

    When every sentence is True under this model. So a set of sentences is consistent iff there is at least one model under which it is consistent. The semantic definition states that a theory is consistent if it has a model, i.e., there exists an interpretation under which all formulas in the theory are true. This is the sense used in traditional Aristotelian logic, although in contemporary mathematical logic the term satisfiable is used instead.

22. [Logic] When can we say that P ⊨ B? P entails B

    Iff P is consistent under interpretation x, then B is true under interpretation x.

23. [Logic] What is "this set of sentences is consistent" short for?

    This set of sentences is consistent with at least one model.

24. [Logic] Why propositional logic is monotonic?

    Because adding sentences to its knowledge base increases (or keeps) the number of sentences entailed by it.

25. [Logic: reasoning process] Write as the A → B statement: soundness completeness.

    soundness - can be reasoned to → is entailed completeness - is entailed → can be reasoned to.

26. [Logic: Propositional Logic] What's a unit resolution rule in Propositional Logic?

    Allows us to remove one of the disjuncts.

27. [Logic: Predicate Logic] Name 2 interesting rules of the reasoning process in Predicate Logic. Reasoning process in Predicate Logic is complex, however 2 rules are particularly interesting.

    Universal Instantiation - choose the constant from the knowledge base. ∀x A(x) entails A(Old)Existential Instantiation - introduce the new constant. ∃x C(x) entails C(New)