Machine learning methods: from linear models to neural networks

Discovery mode

The linear model and the neural network principle and the goal are basically consistent, the difference manifests in the derivation link. If you are familiar with the linear model, the neural network will be well understood, the model is actually a function from input to output, we want to use these models to find patterns in the data, to discover the existence of the function dependencies, of course, if the data itself exists such a function dependency. There are many types of datasets, which may be structured, such as tables in relational databases or unstructured, such as text. We are now considering a typical data set, which is a relational table, each record represents an entity, such as a flower, the relationship table has n fields, the first n-1 field is the various characteristics of the entity, such as the length and width of the petals; the last field is the label of the entity, and the label is a discrete value. For example, if a record represents a plum blossom, then its label 0, if it is an orchid, then its label is 1, of course, it can be other values. We use x to represent the first n-1 fields of each record, and Y to represent each record's label.

Now our goal is to find a pattern, that is, the function dependency F from the feature to the label, so that there is f (x) ≈y for as many records as possible.

\begin{equation} \int_0^\infty \frac{x^3}{e^x-1}\,dx = \frac{\pi^4}{15} \label{eq:sample} \end{equation}

Logistic regression model

Linear regression model

Neural network

　　

Machine learning methods: from linear models to neural networks