Summary of Week 3 -- Capsule 1 -- Nearest Neighbor

00:00:00 - 00:15:00

The "nearest neighbor" method of machine learning is a non-parametric approach that does not require fitting parameters using training data. The only hyper parameter that you have to set is the number of neighbors. The concept is illustrated with a brief example of using it to predict the price of a new test house.

• 00:00:00 In this week's video, we'll be introducing the nearest neighbor model, which belongs to the non-parametric category of supervised learning models. This model works by looking at the closest point to a query instance, and then predicting the class of the instance based on its nearest neighbor.
• 00:05:00 In this video, the presenter demonstrates how to use nearest neighbor classification to determine the class of a new point. They compare two approaches, one where k is equal to three and one where k is equal to one. The decision boundary is smoother when k is higher.
• 00:10:00 This video explains the properties of nearest neighbor classifiers, including that they are non-parametric, require storage of the entire data set, and can be expensive to search. It also notes that if the training data is continuous, the Euclidean distance function can be used, while other distance functions can be used when using non-continuous data.
• 00:15:00 The "nearest neighbor" method of machine learning is a non-parametric approach that does not require fitting parameters using training data. The only hyper parameter that you have to set is the number of neighbors. The concept is illustrated with a brief example of using it to predict the price of a new test house.