Basic concepts of machine learning

1 why the random gradient descent method can work?

The explanation of https://www.zhihu.com/question/27012077 in the Li Wenzhe of the respondents
2 The advantages of the stochastic gradient descent method? (1) speed up the training speed (2) noise can make out of the local optimal 3 trade-offs variance and deviation : deviations reflect the accuracy of the model (the degree to which the training data is matched), and the variance reflects the stability of the model (the ability to generalize the test data). The more complex the model, the smaller the deviation, the greater the variance. 4 methods to reduce overfitting : (1) Reduce the number of features (2) Increase the amount of data (3) introduce a regular term

Basic concepts of machine learning