introduction
Full book chapters
0101 Introduction to basic concepts
1.1 Introduction
What problem does machine learning solve?
- By means of calculation, use experience to improve the performance of the system itself
- With data
- Through some kind of learning algorithm
- Get the model
- Make predictions
1.2 Basic terminology
With data
- Data set , for example 100 watermelons
- Sample , for example 1 watermelon
- Feature vector The
feature vector becomes the sample space.
If the feature vector has color, size, and amplitude, then the dimension is three-dimensional, and the sample space is like a three-dimensional coordinate system. - Attribute , one of the feature vectors, such as color
Through some kind of learning algorithm
- The process of learning a model from data is called learning or training . This process is completed by executing a learning algorithm
Get the model
Supervised learning | Unsupervised learning |
---|---|
The training data is marked with correct answers | The training data is unlabeled and there is no clear answer |
Includes classification and regression | Include clustering |
-
Classification : the discrete value to be predicted, such as "good melon", "bad melon"
- Two categories : only involves two categories-positive and negative, such as judging whether a melon is ripe or not, should it be picked?
- Multi-category : The results involve multiple categories, for example, there are 3 types of watermelons: Black Beauty, Little Landmine, and Te Xiaofeng, which one to buy
-
Regression : What I want to predict is a continuous value, such as watermelon sweetness 0.95, 0.37
- You can also predict the price of watermelon in various time periods next year based on the price trend of watermelon in previous years
- You can also predict the price of watermelon in various time periods next year based on the price trend of watermelon in previous years
-
Clustering : We don’t know how many categories to divide, the machine divides it by itself
- Each group is called a " cluster " (cluster)
Make predictions
- Test: After learning the model, the process of using it to make predictions
- Test sample: the predicted sample
- Generalization ability: The ability of the learned model to apply to new samples. A model with strong generalization ability can be well applied to the real sample space
0102 Hypothetical space induction preference
1.3 Hypothetical space
Scientific reasoning
- Induction: special to general
- Deduction: general to special
1.4 Inductive Preference
The simpler the model, the more it reflects the essence,
just like a mathematical formula