Talk about machine learning 1. Error small vs probability probability max (3)

Encounter a new problem

So far, at least a few issues need to be addressed:

What to do if the analytic solution cannot be solved. The answer is: Iterative Search solver (optimization problem)

If the number of samples is particularly large, how to deal with it. The answer is: recursive solver (online learning optimization problem)

What if the error is not a normal distribution? The answer is: generalized linear regression

If the dimension of this matrix (XXT) −1 {\left ({x{x^t}} \right) ^{-1}} is particularly large, how to handle it. The answer is: Data dimensionality reduction

If the model is nonlinear, how to deal with it. The answer is: non-linear model

The above process is all the frequency of the point of view, how the Bayesian faction do it. The answer is: Build the model

What is the relationship between linear regression and deep learning? The answer is: when I summarize, haha

So our machine learning journey is officially started.