Watermelon Book Reading Notes (1)-Introduction

Summary of all notes: "Machine Learning" Watermelon Book-Summary of reading notes

1. Introduction

Machine learning is dedicated to the study of how to use experience to improve the performance of the system itself through computational means.

2. Terminology

Data set: A collection of records describing the characteristics of a single sample. The description of an event or object in each record is called an "example" or "sample", and items that reflect the performance or nature of the event or object in a certain aspect are called "attributes" or "features."

…(I found that there is nothing to say about this area. If you use more of these concepts, you can understand what you are talking about. If you really don’t understand, just read the book)

Three, hypothetical space

Induction and deduction are the two basic methods of scientific reasoning. The former is the process of "generalization" from special to general, that is, from specific facts to general laws; the latter is the process of "specialization" from general to special, that is, deduced from basic principles The specific situation.

We can regard the learning process as a process of searching in a space composed of all hypotheses. The search goal is to find a hypothesis that "fits" with the training set, that is, the hypothesis that can be judged correctly in the training set.

Fourth, inductive preferences

For a specific algorithm, we need to choose a model. At this time, the algorithm preference we choose will play a key role. Any effective machine learning algorithm must have its inductive preference, otherwise it will be confused by hypotheses in the hypothesis space that seem to be "equivalent" on the training set, and will not produce definite learning results.

"Occam's razor" (Occam's razor) is a commonly used and the most basic principle in natural science research, that is, "if there are multiple hypotheses consistent with observations, choose the simplest one." It is helpful to reduce the effect of the model brought by overfitting on the training set better than the test set.

5. Development process, application status and reading materials

1. Data mining: discover the relationship between data
2. Computer vision: understand the world like a human
3. Natural language processing: understand text like a human
4. Robot decision-making: have the ability to make decisions like humans

Others, look at the textbook, read it further...

Next Chapter Portal: Watermelon Book Reading Notes (2)-Model Evaluation and Selection