Deep learning (001) - Deep Learning Introduction

I. Introduction:
1. branch of machine learning, artificial neural networks (Artificial Neural Network, ANN) based

Second, with the machine learning differences:
1. Machine Learning manually automatic feature extraction feature extraction depth study
2. Machine learning, data less, relative ineffective depth study, data from multiple, relatively better

Third, the artificial neural network:
1. (Artificial Neural Network, ANN) mathematical model to mimic the structure and function of biological neural networks (center of the brain) is used to solve the problem of function approximation or estimate awakened better

IV neurons:
1. The basic neural network, connected to each other to form the neural network
2. T = f (W ^ TA + b)

Fifth, the single-layer neural network:

Sixth, Perceptron:

Seven, multilayer neural network:

Eight, activation function:
1. Role: to increase the capacity to improve the robustness of nonlinear segmentation model (robustness fit another wave of data capacity) to alleviate the problem disappear gradient accelerating convergence of the model (model train faster), etc.
2. Example: Perceptron Add half of the incomplete half-line activation function makes a straight line increase the accuracy of distinguishing bowing (an example can be understood)

Other:
Linear condition: