TY - GEN
T1 - Concept learning as motor program induction
T2 - 34th Annual Meeting of the Cognitive Science Society: Building Bridges Across Cognitive Sciences Around the World, CogSci 2012
AU - Lake, Brenden M.
AU - Salakhutdinov, Ruslan
AU - Tenenbaum, Joshua B.
N1 - Funding Information:
We gratefully acknowledge Jason Gross for developing the experimental interface and for collecting the data. We also thank John McCoy for helpful comments, and Dan Ellis for making his DTW code available.
Publisher Copyright:
© CogSci 2012.All rights reserved.
PY - 2012
Y1 - 2012
N2 - Human concept learning is particularly impressive in two respects: the internal structure of concepts can be representationally rich, and yet the very same concepts can also be learned from just a few examples. Several decades of research have dramatically advanced our understanding of these two aspects of concepts. While the richness and speed of concept learning are most often studied in isolation, the power of human concepts may be best explained through their synthesis. This paper presents a large-scale empirical study of one-shot concept learning, suggesting that rich generative knowledge in the form of a motor program can be induced from just a single example of a novel concept. Participants were asked to draw novel handwritten characters given a reference form, and we recorded the motor data used for production. Multiple drawers of the same character not only produced visually similar drawings, but they also showed a striking correspondence in their strokes, as measured by their number, shape, order, and direction. This suggests that participants can infer a rich motor-based concept from a single example. We also show that the motor programs induced by individual subjects provide a powerful basis for one-shot classification, yielding far higher accuracy than state-of-the-art pattern recognition methods based on just the visual form.
AB - Human concept learning is particularly impressive in two respects: the internal structure of concepts can be representationally rich, and yet the very same concepts can also be learned from just a few examples. Several decades of research have dramatically advanced our understanding of these two aspects of concepts. While the richness and speed of concept learning are most often studied in isolation, the power of human concepts may be best explained through their synthesis. This paper presents a large-scale empirical study of one-shot concept learning, suggesting that rich generative knowledge in the form of a motor program can be induced from just a single example of a novel concept. Participants were asked to draw novel handwritten characters given a reference form, and we recorded the motor data used for production. Multiple drawers of the same character not only produced visually similar drawings, but they also showed a striking correspondence in their strokes, as measured by their number, shape, order, and direction. This suggests that participants can infer a rich motor-based concept from a single example. We also show that the motor programs induced by individual subjects provide a powerful basis for one-shot classification, yielding far higher accuracy than state-of-the-art pattern recognition methods based on just the visual form.
KW - concept learning
KW - one-shot learning
KW - program induction
KW - structured representations
UR - http://www.scopus.com/inward/record.url?scp=84898967430&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=84898967430&partnerID=8YFLogxK
M3 - Conference contribution
AN - SCOPUS:84898967430
T3 - Building Bridges Across Cognitive Sciences Around the World - Proceedings of the 34th Annual Meeting of the Cognitive Science Society, CogSci 2012
SP - 659
EP - 664
BT - Building Bridges Across Cognitive Sciences Around the World - Proceedings of the 34th Annual Meeting of the Cognitive Science Society, CogSci 2012
A2 - Miyake, Naomi
A2 - Peebles, David
A2 - Cooper, Richard P.
PB - The Cognitive Science Society
Y2 - 1 August 2012 through 4 August 2012
ER -