I am finding it difficult to differentiate Multi-class, Multi-label and multi-instance in Machine learning
Example 1:
Predictor variables : Assignment 1 score, Mid term exam score, and prior GPA for particular course (mid of the semester )
Output variable: Final Grade (could be A or B or C or D ). Here you are attempting to classify /predict student grade which belongs to more than two classes but one at a time. A student grade should be A or B or C or D. So, this is called Multi-class classification.
Example 2: Predictor variables: Book name; Book author; age group-can read Abstract (not more than 50 characters)
Output variable: Type of book (could be Comedy / Fiction/ Horror/ Romance/ Adventure./ or combinations of these types)
Here you are attempting to classify /predict book type which belongs to more than two classes (many types) but it may have more than one labels at a time. That is the book can be classified as Comedy and Fiction / Adventure, Horror and Fiction / Mystery only. So, this is called Multi-label classification.
Sorry, I have not used Multi-instance in machine learning so far. Probably someone will come up with examples.
- Badges
- Science topic