Unsupervised and supervised methods to understand there are many, can be understood from the following main aspects:
1, no difference between supervised and unsupervised learning that a value-free education, a teaching value. But personally I think that the difference between them is that the general is the use of unsupervised learning clustering algorithms to classify different samples. The supervised learning in general is the use of teaching error value and the actual output value generated by the error back-propagation network to modify the weights to complete the correction. But unsupervised learning does not modify the weights back-propagation operation.
2, all tagged samples / sample all untagged ··· PS: section labeled semi-supervised learning
3, the training set input the output is supervised, including all regression algorithm classification algorithm, such as linear regression, decision trees, neural networks, KNN, SVM and so on; the training set not only the input output is unsupervised, including all clusters algorithms, such as k-means, PCA, GMM , etc.
There supervised learning: the concept of training samples have labeled (classification) of learning, data outside the training sample set as labeled (classified) forecast. Here, all tags (classification) is known. Therefore, the ambiguity of training samples is low. Unsupervised Learning: no concept of training samples labeled (classification) is to learn, to discover the structural knowledge of the training sample set. Here, all tags (classification) is unknown. Therefore, the ambiguity of training samples is high. Clustering is a typical unsupervised learning
