Statistical Learning
If a system can improve its performance by performing a process, this is learning.
Data + Model -> Prediction
Object - Data
Numbers, text, images, video, audio and their combinations
- Extracting data features -> Abstract out a data type -> Discovery knowledge in data -> Analysis and forecasting data
- The basic assumption about data is that the data with a common attribute, such as data in some database, have certain statistical laws. In this case, we can process data by means of probability and statistics.
- Using random variables to describe the features in data - continuous random variable and discrete random variable
- Using probability distribution to describe the statistical law of data
Purpose - Analysis & Predication
by building a probabilistic statistical model
The general purpose is to determine which model to learn and how to learn the model, considering the learning efficiency, such that the model can make accurate prediction and analyses.
Method
Supervised learning, unsupervised learning, semi-supervised learning and reinforcement learning
- Steps to implement statistical learning methods
- Get a finite training set
- Determine the hypothetical space that contains all possible models
- Determine the criteria for model selection (strategy)
- Implement the algorithm for finding the optimal model
- Using the optimal model predict and analyze new data
