Domains of values of things in the real world:
A real number. This subsumes boolean (0,1) and probability 0...1.
Angle / point on a circle. This is qualitatively different from a real number, because 359 degrees is close to 0.
Point on a sphere.
Rotation in three dimensions.
Tuples of the above.
A tuple of real numbers on the subspace such that the tuple sums to unity. Is the subspace constraint useful? Far more complicated subspaces are possible, but do we care about them?
Discrete categories. However these can be modeled by making each category item a Bernoulli trial and constraining the categories to sum to unity as above.
Unions (probably tagged unions) of two or more types. This permits the Maybe type, allowing something to have no value.
An unbounded list of items of one type.
Inspiration is features and labels for the classification problem in machine learning.
No comments :
Post a Comment