AI Exam Practice Questions: T3 Topics
Uploaded by
helloboy12361AI Exam Practice Questions: T3 Topics
Uploaded by
helloboy12361Common questions
Powered by AIConflict resolution in a rule-based system can be managed using strategies such as specificity ordering, recency ordering, and priority ordering. Specificity ordering chooses the most specific applicable rule. Recency ordering selects the rule that uses the most recently added facts. Priority ordering applies predefined priorities to the rules. These strategies help determine which rule to fire when multiple rules match .
In A* algorithm, the Manhattan heuristic calculates f-values based on grid-based distance (sum of absolute differences in horizontal and vertical distances), whereas the Euclidean heuristic uses straight-line distance (direct spatial distance). The Manhattan heuristic is often more efficient in grid-based environments due to its simplicity and ties to the grid structure, while the Euclidean heuristic can provide more accurate estimates in open continuous spaces, potentially leading to more optimal paths but at the cost of increased computation .
An example of a heuristic function that consistently underestimates path costs is the 'zero heuristic', which assigns a zero estimated distance to the goal from any node. A more practical underestimating heuristic might use the minimum movement to the goal along any dimension in a grid-based system. Since it is admissible (never overestimates), it ensures A*'s optimality by not disregarding any potentially optimal paths, thus prompting the examination of all viable paths before concluding .
Semantic networks allow for intuitive visualization and flexible representation of knowledge structures through nodes and links, making inheritance and relationship capture straightforward. However, they lack the procedural power of rule-based systems for complex inferencing and can become cumbersome with large networks. Rule-based systems offer precise control and are suitable for tasks requiring explicit procedural logic but can be less flexible in representing non-linear relationships .
Slot inheritance in frame-based systems allows child frames to inherit attributes and values from parent frames, thus minimizing redundancy. For instance, the 'Fruit' frame might have slots for 'color' and 'taste', with default values. The 'Apple' frame inherits these slots but can override them with specific values, like setting 'color' to 'red'. This process of inheritance streamlines knowledge representation by sharing common characteristics among related objects .
Move ordering significantly affects the performance of alpha-beta pruning by reducing the tree's search space. By ordering moves from most promising to least (usually determined by heuristics or historical data), unnecessary branches can be pruned earlier, increasing efficiency. For example, if the best move is ordered first, alpha and beta cutoffs occur sooner, reducing the number of nodes evaluated. This ordering reduces computation time and improves scalability in deep game trees .
A* might be preferred over Greedy algorithms even if the latter examines fewer nodes because A* guarantees finding the shortest path due to its heuristic optimality condition, which combines both actual cost to reach a node (g) and estimated cost to the goal (h). In contrast, Greedy search focuses solely on the estimated cost to the goal, which can lead to suboptimal paths. A* provides consistent optimal solutions and balances exploration and exploitation better than Greedy, making it generally more reliable .
To create a rule-based irrigation controller, define rules based on time and soil moisture. For example: Rule 1: IF time is between 6 AM and 8 AM AND soil_moisture is below 50%, THEN start irrigation. Rule 2: IF time is between 6 PM and 8 PM AND soil_moisture is below 40%, THEN start irrigation. These rules ensure the system reacts at optimal times with respect to moisture levels, integrating time-based and environmental cues for efficient irrigation control .
Constructing a semantic network involves creating nodes for concepts like 'Whale', 'Mammal', and 'Animal'. Using IS_A links, you connect 'Whale' to 'Mammal', and 'Mammal' to 'Animal'. The inference that 'Whale is an Animal' follows the path of these IS_A links. As semantic networks naturally support inheritance through these hierarchical links, the inference engine can traverse the network to deduce properties from parent classes, demonstrating effective hierarchical reasoning .
A single perceptron cannot solve the XOR logic gate because XOR is not linearly separable. Perceptrons can only classify linearly separable data. To solve XOR, a multilayer perceptron (MLP) with at least one hidden layer is required. The hidden layer allows the network to transform the input space into a new space where the classes become linearly separable .