1 / 23

Category Generation

Category Generation. Alan Jern Charles Kemp Carnegie Mellon University. Category generation: Introduction. Observation: people can imagine and create new objects. Cobb salad. Spork. Griffin. T. Ward (1994). Category generation. Outline. Define category generation Modeling approach

yori
Télécharger la présentation

Category Generation

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. Category Generation Alan Jern Charles Kemp Carnegie Mellon University

  2. Category generation: Introduction • Observation: people can imagine and create new objects Cobb salad Spork Griffin T. Ward (1994) Category generation

  3. Outline • Define category generation • Modeling approach • Experiment

  4. Category generation • Can occur at any level in a taxonomy Category level Exemplar level

  5. Category generation This talk

  6. Training Caesar Greek Classification x1 x2 x3 x4 x5 x6

  7. Test Caesar Greek ? Classification x1 x2 x3 x4 x5 x6 xnew

  8. Caesar Greek ? Classification x1 x2 x3 x4 x5 x6 xnew Training Caesar Greek Category generation x1 x2 x3 x4 x5 x6

  9. Caesar Greek ? Classification x1 x2 x3 x4 x5 x6 xnew Test Caesar Greek xnew Category generation ? x1 x2 x3 x4 x5 x6

  10. Past modeling approaches • Classification • Prototype model (Reed, 1972) • GCM (Nosofsky, 1985) • Rational model (Anderson, 1991) • ALCOVE (Kruschke, 1992) • SUSTAIN (Love et al., 2004) • Category generation • ?

  11. Test Caesar Greek ? Classification x1 x2 x3 x4 x5 x6 xnew Caesar Greek xnew Category generation ? x1 x2 x3 x4 x5 x6

  12. Our approach • Bayesian • Learn category distribution • Sample from that distribution

  13. A category generation task … M M W G R G K B K X X R

  14. An exemplar generation task Pieces Structure G K N B V S MX HF WR MX G K Slot 1 Slot 1 Slot 2 W M M Slot 2 N G G K B K X X R cf. Fiser & Aslin, 2001

  15. Task model Structure ? G K N B V S MX HF WR MX G K Slot 1 Slot 2 ? Slots Slot 1 M W W M Slot 2 G N N G B K B K R X X R

  16. Experimental Design • 18 CMU undergraduates • 3 structures (3 conditions)

  17. Materials Slot 1 Z N Q J V S W K DB Seen combination RL Slot 2 XM HF

  18. Procedure

  19. Results No item generated more than twice Humans Model 58% of responses (Hypothesis space: 1820 possible responses)

  20. Results • Classification task • Rate likelihood of new instances Valid All Distractors 1 Seen Pairing 3 Seen Pairings 2 Seen Pairings

  21. Conclusions • People can: • Learn latent category structure • Generate new category members • People are sensitive to frequency differences • Predicted by our probabilistic approach

  22. Conclusions • Category generation is an understudied aspect of human categorization

  23. Thanks • Faye Han • John Anderson • David Rakison Credit: RaynorGanan (ragbag.tumblr.com)

More Related