Maching Learning

Machine learning system type

机器学习的类型划分主要基于以下几点：

• Whether or not they are trained with human supervision
  • supervised: regression prediction and classification（回归预测和分类）
  • unsupervised：clustering（聚类）
  • semisupervised
  • Reinforcement Learning
• Whether or not they can learn incrementally on the fly
  • online
  • batch learning
• Whether they work by simply comparing new data points to known data points, or instead detect patterns in the training data and build a predictive model, much like scientists do
  • instance-based
  • model-based learning

Reference Link:https://blog.youkuaiyun.com/huang1024rui/article/details/49735113

Pattern Recognization Machine learning

what does PRML do?

• Classification: Assign each input to one of a given set of classes
• Regression: assign a real-valued output to each input
• Sequence Prediction: Do classification or regressioin to each member of a sequence of values
• Decision making: assign a series of actions to achieve the best overall reward

Two stage of machine learning

• Training of learning phase
  • Training data
  • determing f(x)
• Test phase
  • Testing: classification/predication/decoding…
  • Generalization problem

Key Element of Machine learning

• Type of machine learning
  • example : blind source separation type is ??
• Data
  • input (observation feature)
  • output(label)\
• model
  • Parametric or non-parametric
  • Form of model (what are the parameter)
• Criterion
  • what’s the meaning of being optimal

Generative VS Discriminative

About the difference, may the link:
https://blog.youkuaiyun.com/zouxy09/article/details/8195017 will help.

Training Model

Linear

Gradient Descent(GD)

• batch GD
• Mini-batch GD
• Stochastic GD

Polynomial Regression

SVM(support vector machine)

svm

Decision Tree

Ensemble learning and Random Forests

Neural network

Activate function

DNN(deep neural nets)

CNN(convolutional neural nets)

pros:

• minimize model parameter numbers
• enhance generilize ability: able to learn the intrinstic of data, and can be transfer the ablity to new group of data