Learning notes for "Machine Learning Practice": Implementation of k-Nearest Neighbor algorithms, and "Machine Learning Practice" k-

Learning notes for "Machine Learning Practice": Implementation of k-Nearest Neighbor algorithms, and "Machine Learning Practice" k-

The main learning and research tasks of the last semester were pattern recognition, signal theory, and image processing. In fact, these fields have more or less intersection with machine learning. As a result, I continue to read machine learning and watch the machine learning courses at Stanford University. In this process, because of the future project requirements of the research group, we need to contact Python. Therefore, we chose the "Machine Learning Practice" book and learn together with the reference materials and videos. In fact, the book's theoretical research is not deep enough. It can only be a tool for practicing Python and verifying some famous machine learning algorithms.

Before introducing k-Nearest Neighbor algorithms, perform simple classification and sorting on machine learning algorithms. In simple terms, machine learning is divided into two main categories: supervised learning) and unsupervised learning ). Supervised Learning can be divided into two categories: classification .) and regression (regression), a classification task is to classify a sample into a certain known category, and the category information of each sample needs to be given during training, for example, face recognition, behavior recognition, and target detection are all classified. The regression task is to predict a value, such as the given housing market data (area, location, and other sample information) to predict the price trend. Unsupervised learning can also be divided into two types: clustering and density estimation. clustering aggregates a pile of data into a weak stem group without category information; density Estimation is to estimate the statistical parameter information of a pile of data to describe the data, such as the RBM of deep learning.

As the first chapter, the author introduces a simple and easy-to-understand k-Nearest Neighbor (kNN) algorithm. This algorithm is used together with the Parzen window Estimation in pattern recognition. It is a non-parameter estimation method. This type of method is used to process any form of probability distribution without prior consideration of the parameter form of probability density.