First, what is machine learning? What is machine learning?
Personal understanding: machine learning is the input data set model, learning through existing data, based on previous learning to new data to predict when an incoming new data.
Machine learning method based on the presence of certain data to generate a certain algorithm model law, also known as learning algorithm.
Second, the classification of machine learning what?
Supervised learning, semi-supervised learning, unsupervised learning, reinforcement learning
1. supervised learning
1.1 Definitions
Label data, characteristic
The general idea: modeling for the association between the characteristic data and the label, the model is determined, a learning algorithm may be applied to the new data.
1.2 learning process and task
Known input and labeling features, training model, enter a new predicted characteristics resulting label, evaluate the effect of view of the model (predicted is known label tag comparison)
Tasks are divided into: Category classification and regression regression
Data type label: label classification task is discrete data; the return label is a continuous task of Data
2. semi-supervised learning
Have ranged between supervised learning and unsupervised learning, applied to the data is incomplete
3. unsupervised learning
3.1 Definitions
Data without any labels
A learning algorithm for data modeling feature does not have any tags (data yourself Introduce yourself)
3.2 learning task
Clustering clustering and dimension reduction dimensionality reduction
Clustering algorithms: data partitioning category
Dimensionality reduction algorithm: The law performance with compact form
# Supervised learning and unsupervised learning of inductive
Supervised learning: data characteristic, label classification, regression algorithm
Unsupervised Learning: no tag data clustering, dimension reduction algorithm
4. Reinforcement Learning
To 'trial and error' way of learning, based on their real-time status and actions to adapt to changes in the environment, constantly evaluate the temptation to achieve the best results.
Third, the learning algorithm regular contact
1. supervised learning
k-nearest neighbor (KNN), decision trees, naive Bayes, return to the family, etc.
2. Unsupervised Learning
Clustering, dimensionality reduction (principal components PCA), etc.
These are some of my knowledge of the algorithm are summarized, is not comprehensive enough, please understand.