Final Exam.docx

Chapter 6 Review: In your own words, define the core differences between supervised and unsupervised learning? Essay Supervised learning implies that a teacher is present to define when an answer is correct or incorrect. Data with classes is one of the most common ways of presenting data to superv...

Chapter 6 Review: In your own words, define the core differences between supervised and unsupervised learning? Essay Supervised learning implies that a teacher is present to define when an answer is correct or incorrect. Data with classes is one of the most common ways of presenting data to supervised learning algorithms. Unsupervised learning algorithms learn without a teacher, and simply exploit patterns or relationships in the data. What is a common application of decision tree learning? Essay One of the most popular uses of decision tree learning is in the field of data mining. This application desires to understand the relationships of data as well as extract useful information from raw data. Decision tree learning can do this by segmenting large amounts of data into more compact decision trees and identifying the key aspects of the data that can be used to segment it. Decision tree learning is also very fast, making it suitable for dealing with large data sets (as is typical in data mining applications). Define entropy and its application to decision tree learning. What can be inferred if entropy is zero or one? Multiple Choice Entropy in decision tree learning is used to find out how much information is behind a given feature (based upon information gain from information theory). If entropy is zero, the feature cannot be used to segment the data because the feature belongs to one class. When entropy is one, the features are evenly split making it a useful feature to segment the data. What issues can result from creating decision trees from training sets that are too small or too large? Large decision trees, or those created from small amounts of data tend to over-generalize. In other words, the tree matches the data instead of providing a generalization of it. This problem can be solved by pruning (or minimizing) the training data to avoid over-fitting. Define the queuing structures used in DFS and BFS and explain why each uses their particular style. Depth-First Search (DFS) uses a stack (or LIFO) structure to support searching branches to their greatest depth before backtracking to previously unexplored branches. Breadth-First Search (BFS) uses a queue (or FIFO) to support search of most recently unexplored nodes first. DFS Versus BFS DFS, Depth First Search, is an edge-based technique. It uses the Stack data structure and performs two stages, first visited vertices are pushed into the stack, and second if there are no vertices then visited vertices are popped.  Breadth-First Search, is a vertex-based technique for finding the shortest path in the graph. It uses a Queue data structure that follows first in first out. In BFS, one vertex is selected at a time when it is visited and marked then its adjacent are visited and stored in the queue. It is slower than DFS.  Minimax can search to the leaves of the tree, or to a predefined depth. What are the consequences of ending search at a predefined depth? Searching a shallow depth of a tree results in what’s called the “horizon effect.” This means that the moves viewed so far may look optimal, but once the search sees over the horizon, it may find that the moves have resulted in a losing strategy.