RMG 300 Lecture Notes - Lecture 12: Unsupervised Learning, Computer Vision, Supervised Learning
11 Dec 2018
School
Department
Course
Professor
Document Summary
Database basics: traditional database organizes data in a hierarchy, a field is the smallest element. Fields are within a record: records are organized in a data file, database is a collection of data files. Machine learning: data preparation, algorithms, sequence of steps to solve a problem, easily repeated to solve similar problems, goals, good data. Supervised learning 70: trained using labeled examples. Unsupervised learning: data exploration, looking for patterns. Neural networks: trial and error process, data fed in model works out a solution, data in training mode must be labelled with metadata, natural language processing, speech recognition, self driving cars. Issues: training set too small or not representative, biases. Internet of things: most of the data is unstructured. Unsupervised learning: unlabelled data most data in the world, the more data; the more accurate, search: comparing documents, images or sounds to similar items, detecting anomalies. Storage for huge amounts of data - any kind. Fault tolerance redirected if a node fails.
