Introduction_Lecture1.pdfML講義關於PATTERN RECOGNITION

文件名称: Introduction_Lecture1.pdf

所属分类: 讲义

开发工具:

文件大小: 3mb

下载次数: 0

上传时间: 2019-10-20

提供者: kore****

下载 (3mb)

不能下载？报告错误

详细说明：ML講義關於PATTERN RECOGNITION AND MACHINE LEARNINGWhat is machine learning? o Automatic discovery of regularities in data Algorithms and techniques that allow computers to learn". the major focus is to extract information from data automatically, by computational and statistical methods e Applications: natural language processing, search engines, medical diagnosis, bioinformatics, stock market analysis, game playing and robot locomotion ehttp:/len.wikipediaorg/wiki/machinelearning Relation to"intelligence Learning is fundamental characteristic of human intelligence To learn is to change for the better o One way to measure change is in terms of behavior of organism in new but similar situations Generalization is key: it is easy to learn by heart, difficult to learn general-purpose strategies o Useful distinction innate versus acquired knowledge (for us: priors versus data) elated publications Conference Journal Machine Learning(ML) ICML Journal of Machine Learning KDD Research NIPS Annals of statistics IJCNN Data Mining and Knowledge AIML Discovery IEEE-KDE IJCAI IEEE-PAMI COLT Artificial Intelligence CVPR ournal of artificial ICCV Intelligence research ECCV Computational Intelligence Neural Computation IEEE-NN Research, Information and Computation Q1. What is machine learning? First, we define learning as using past experiences to improve future performance For machines experiences come in the form of data So, machine learning is the design of algorithms that can"improve their performance"(given some quantifiable measure) with experience Machine Learning Problems Not Ml problems: sorting, Traveling 命 Salesman. 3-Sat etc ML Problems: hard to formalize, but human expert can provide examples feedback Computer needs to learn from eedback Is there a sign of cancer in this fMRI scan What will the dow jones be tomorrow Getting a robot to ride a unicycle Q2. Why use learning Often it is too difficult to design a set of rules " by hand" Machine learning is about 1. automatically extracting relevant information from data 2. applying it to analyze new data Q3. When should we use learning 1. A pattern exists 2. We cannot pin it down mathematically 3. We have data on it Face detection Does a patch of an image represent a face? B=2 Components of a ML Problem Task: What are we trying to do? This is important to be sure that our example (training data is useful Experience: What data do we provide the algorithm? This defines the input to the learning system and the data on which it bases its decisions. It also defines the representation of the input and output? Performance Metrics: How do we measure how well the system is doing? This gives us an objective measure to judge the learning process. It also allows comparison between competing methods

(系统自动生成,下载前可以参看下载内容)