Theory-Based Induction
Lists
Only you see this
Abstract
We show how an abstract domain theory may be incorporated into a rational statistical model of induction, giving a more principled basis for the model's assumptions, greater explanatory power, and closer contact with processing-level models. In particular, we develop a Bayesian model of category-based induction that is founded on a theory of how biological properties are distributed across classes of biological kinds. We argue that our new approach is much more principled than previous Bayesian and similarity-based models of inductive inference about animal categories. We also show that our new approach performs better than these previous models over a collection of four data sets, one of which is new.
BibTeX
@inproceedings{kemp2003theory,
author = {Charles Kemp and Joshua B. Tenenbaum},
title = {{Theory-Based Induction}},
booktitle = {eScholarship},
year = {2003},
}Loading related content...