Hypothesis spaces

Hypothesis spaces

par FILIPPO FANTINATO,
Nombre de réponses : 0

Binary classification

A binary classification task means to split the instance space in two sets. Some examples can be:

  • to classify whether an image represents an apple or not;
  • to classify whether the cat in an image is big or not;
  • to classify whether an email is spam or not.

Regarding the first example, the instance space can be split by assigning +1 to the images representing an apple and -1 otherwise and then the instance space can be split via a line, a circle or a diamond.

Multi-class classification

A multi-class classification task means to split the instance space in more sets, each one representing a class. Some examples can be:

  • to classify what words a user is pronouncing;
  • to classify what species an animal belongs to;
  • to classify what categories an email belongs to.

The instance space can be split by several straight lines, curved lines or geometric figures, where the portions of space represent the classes. 

Regression tasks

A regression task means to find the mapping function to map the input variable x to the continuous output variable y. Some examples can be:

  • weather forecasting;
  • sales forecasting;
  • cash forecasting.

The hypothesis space depends on the degree of polynomial we choose.