Exercise: Identifying a Foo
This example is adapted from the 'arch' example in the the following book:
Winston, P. (1993). Artificial Intelligence (3rd ed.). Addision-Wesley.
In saying 'Yes', you're suggesting a more general model of foo: the object is a foo if it has any shape on top of its supports.
A more specialized model would say that the shape on top must be a brick, which would mean that this is not a foo. We haven't seen enough examples to know which model is correct, though, so we're having trouble settling in to a level of specialization or generalization. That will be the main topic of this lesson: how to use new examples to help specialize and generalize an agent's model of a concept.
Incremental Concept Learning
This algorithm is adapted from the following book:
Winston, P. (1993). Artificial Intelligence (3rd ed.). Addision-Wesley.
eg, Child A has a black cat, then A goes to a friend's house and see an orange cat (is told that the orange cat is a cat) => doesnt fit into the current definition of cat => generalize the concept of "cat"
eg. Child B has a dog => furry animal => goes to firend's house and see a cat. B is told the cat is not a dog => specilize the concept of "dog"
Variabilization
Generalization to Ignore Features
Specialization to Require Features
Specialization to Exclude Features
Generalization to Abstract Features
Generalization with Background Knowledge
An Alternative Visualization
Winston Chapter 16, pages 349-358
Heuristics for Concept Learning
Final Concept of a Foo
