Internal model configuration variables, for example: in the neural network weights w and bias B, SVM support vector coefficients of the linear regression of a logistic regression;
Use the data to estimate (optimization) or the learned;
II. Super model parameters (Hyper-parameters)
Variable outer model set, for example: learning rate of neural networks, the number of iterations, the number of hidden layers, the number of each neural element of C ,, SVM and sigma, k nearest neighbor of K;
Model parameters will not change in the learning process, set by the person directly or select a search algorithm for estimating model parameters.