According to the composition of the data set, machine learning algorithms can be divided into:
- supervised learning
- unsupervised learning
- semi-supervised learning
- reinforcement learning
1 Supervised learning
Definition: The input data is composed of input feature values and target values.
- The output of the function can be a continuous value (called regression),
- Or the output is a finite number of discrete values (called classification).
1.1 Regression problem
For example: predict housing prices, and fit a continuous curve based on the sample set.
1.2 Classification problem
For example: judging benign or malignant based on tumor characteristics, the result is "benign" or "malignant", which is discrete.
2 Unsupervised Learning
Definition: The input data is composed of input feature values, and there is no target value
- The input data is not labeled and there are no definitive outcomes. The sample data category is unknown;
- It is necessary to classify the sample set according to the similarity between samples.
Example:
Supervised and unsupervised algorithm comparison:
3 Semi-supervised learning
Definition: The training set contains both labeled sample data and unlabeled sample data.
Example:
Supervised learning training method:
Semi-supervised learning training method
4 Reinforcement Learning
definition:
The essence is the problem of make decisions, that is, making decisions automatically and making continuous decisions.
Example:
The child wants to walk, but before that, he needs to stand up first, and after standing up, he needs to maintain his balance. Then he has to take a leg first, whether it is the left leg or the right leg, and after taking a step, he needs to take the next step. step.
The child is the agent, who tries to manipulate the environment (the surface on which he walks) by taking actions (i.e. walking), and transitions from one state to another (i.e. each step he takes), when he completes the subtasks of the task (i.e. walking a few steps), the child was rewarded (given the chocolate to eat), and when he could not walk, the chocolate was not given.
It mainly contains five elements: agent, action, reward, environment, observation;
The goal of reinforcement learning is to obtain the most cumulative rewards.
Comparing Supervised Learning and Reinforcement Learning
| supervised learning | reinforcement learning | |
| feedback mapping |
The output is the relationship between them, which can tell the algorithm what kind of input corresponds to what kind of output. |
The output is the feedback reward function to the machine, which is used to judge whether the behavior is good or bad. |
| the feedback time |
Making a bad choice will immediately feed back to the algorithm. | There is a delay in the result feedback, and sometimes it may take many steps to know whether the previous choice of a certain step is good or bad. |
| enter feature |
The inputs are independent and identically distributed. | The input faced is always changing, and every time the algorithm makes an action, it affects the input of the next decision. |
The concept of independent and identical distribution will be explained in the next chapter.
5 Summary
| In |
Out | Purpose | the case | |
| supervised learning |
tagged | have feedback | forecast result | Cat and Dog Classification House Price Forecast |
| Unsupervised learning |
no label | no feedback | discover underlying structures | "Birds of a feather flock together" |
| Semi-Supervised Learning |
some labeled, some unlabeled | have feedback | Reduce the difficulty of data labeling | |
| reinforcement learning |
Decision-making process and incentive system | series of actions | long-term profit maximization | learn to play chess |