one of neural network

 

 

 

map source:https://github.com/microsoft/ai-edu

 

 

• Fundamental Principle

1. inputs: characteristic value that presents attribute of sample.
2. weights: weighted value for each input single. BUT not necessarily their value to 1.
3. bias: deviation measuring learning algorithm for degree of deviation from the expectations of the forecast and actual results, that is, ability to portray fitting of the learning algorithm itself.
4. activative function: simulation of characteristics of biological neurons, after receving inpput by a threshold value simulation of neuron activation and excitement and generate output, to introduce nonlinear neural networks, enhance the ability of neural networks.
5. calculation process: result = (∑w_ix_i + b) * A

• Regression or Fitting

 

 

• Back Propagation; 
• Gradient Descent;
• Loss Function;

 

• FP:

1. Initialization
2. Forward pass
3. Loss function
4. Back propagation
5. Gradient descent
6. Goto 2, loop calculation, To fitting curve.