KBAI: Classification
Imaging that you’ve to tell which type an animal is: is it a bird, a mammal or a reptile? How do you do that? I bet you’ll look at percepts/ properties of the animal to decide: whether the animal has feather or not, whether the animal is warm-blooded or cold-blooded… There are $latex 2^n $ properties combination available, hence there are $latex 2^n $ actions/ decision we can make from those properties.
Formality of concepts:
The more formal a concept is, the more strictly/ explicitly defined it is. Example: Right Angle > Reptitle > Foo > Holiday > Inspirational > Saltiness
Axiomatic Concepts
Conceps that are well-defined as list of necessary and sufficient conditions, such as a Circle: set of point have length R from a center point.
Prototype Concepts
Base concepts defined by a typical example with override properties. Ex: A chair with 3 legs, without back
Exemplar concepts:
Defined by abstraction of instances or exemplars of the concept. Ex: beauty, defined by beauty of a scene, a figure… This don’t even have typical definition.
Top-down classification:
This is use when we have some properties that can differentiate concepts right away. For example, has wing or not can define whether the animal is a bird or a reptile.
Bottom up classification
Look at the nodes properties and abstract it in a bottom-up manner. Useful for things like predict the stock market.
This is an ongoing series of posts where I try to learn a concept by writing blog post about it. The screen shots from this post is taken from the KBAI course in Udacity.