## A more rational model of categorization (2006)

Venue: | Proceedings of the 28th Annual Conference of the Cognitive Science Society |

Citations: | 41 - 16 self |

### BibTeX

@INPROCEEDINGS{Sanborn06amore,

author = {Adam N. Sanborn and Thomas L. Griffiths (tom and Daniel J. Navarro},

title = {A more rational model of categorization},

booktitle = {Proceedings of the 28th Annual Conference of the Cognitive Science Society},

year = {2006}

}

### Years of Citing Articles

### OpenURL

### Abstract

The rational model of categorization (RMC; Anderson, 1990) assumes that categories are learned by clustering similar stimuli together using Bayesian inference. As computing the posterior distribution over all assignments of stimuli to clusters is intractable, an approximation algorithm is used. The original algorithm used in the RMC was an incremental procedure that had no guarantees for the quality of the resulting approximation. Drawing on connections between the RMC and models used in nonparametric Bayesian density estimation, we present two alternative approximation algorithms that are asymptotically correct. Using these algorithms allows the effects of the assumptions of the RMC and the particular inference algorithm to be explored

### Citations

900 | Sequential Monte Carlo Methods in Practice - Doucet, Freitas, et al. - 2000 |

523 |
Arabie P: Comparing Partitions
- Hubert
- 1985
(Show Context)
Citation Context ..., 1000, 0101, 0110, 1011, 1001, 0100 End-Anchored 0100, 0000, 1111, 1011, 0011, 0111, 1000, 1100, 1010, 0001, 0101, 1110, 1001, 0010, 0110, 1101 based on one of the features. The Adjusted Rand Index (=-=Hubert & Arabie, 1985-=-), a standard measure of distance between partitions, was used to find the similarity of the RMC samples to each of the four partitions that split the stimuli along a single feature. The single-featur... |

451 |
A context theory of classification learning
- Medin, Schaffer
- 1978
(Show Context)
Citation Context ...del. Category learning is one of the most extensively studied aspects of human cognition, with computational models that range from strict prototypes (e.g., Reed, 1972) to full exemplar models (e.g., =-=Medin & Schaffer, 1978-=-; Nosofsky, 1986). Recent work has emphasized the “rational” statistical basis of these models (Ashby & Alfonso-Reese, 1995), noting that prototype and exemplar models correspond to different approach... |

427 | Attention, similarity, and the identification-categorization relationship
- Nosofsky
- 1986
(Show Context)
Citation Context ...s one of the most extensively studied aspects of human cognition, with computational models that range from strict prototypes (e.g., Reed, 1972) to full exemplar models (e.g., Medin & Schaffer, 1978; =-=Nosofsky, 1986-=-). Recent work has emphasized the “rational” statistical basis of these models (Ashby & Alfonso-Reese, 1995), noting that prototype and exemplar models correspond to different approaches to the “densi... |

418 |
Mixtures of dirichlet processes with applications to bayesian nonparametric problems
- Antoniak
- 1974
(Show Context)
Citation Context ... densities as possible, so that complex densities can be inferred if they are warranted by the data. The standard model used to solve this problem is called the Dirichlet process mixture model (DPMM; =-=Antoniak, 1974-=-; Neal, 1998). The key idea behind the DPMM is to assume that observations are partitioned into clusters, with the probability of their features depending only on their cluster membership. The prior p... |

375 | Markov Chain Sampling Methods for Dirichlet Process Mixture Models
- Neal
- 2000
(Show Context)
Citation Context ...ssible, so that complex densities can be inferred if they are warranted by the data. The standard model used to solve this problem is called the Dirichlet process mixture model (DPMM; Antoniak, 1974; =-=Neal, 1998-=-). The key idea behind the DPMM is to assume that observations are partitioned into clusters, with the probability of their features depending only on their cluster membership. The prior probability o... |

266 |
Ferguson distributions via polya urn schemes
- Blackwell, MacQueen
- 1973
(Show Context)
Citation Context ...= s s∏ ∏n−1 (nk − 1)! (6) i=0 [α + i] k=1 where α is the concentration parameter of the Dirichlet process. This distribution over partitions can be produced by a simple sequential stochastic process (=-=Blackwell & MacQueen 1973-=-). If observations are assigned to clusters one after another and the probability that observation i + 1 is assigned to cluster k is P (k) = { nk i+α , α i+α , nk > 0 (i.e., k is old) nk = 0 (i.e., k ... |

216 | The Adaptive Nature of Human Categorization - Anderson - 1991 |

127 |
Exchangeability and related topics
- Aldous
- 1985
(Show Context)
Citation Context ...he resulting partition. This distribution has a number of nice properties, including exchangeability: the probability of a partition is unaffected by the order in which the observations are received (=-=Aldous, 1985-=-). It should be apparent from our description of the DPMM that it is similar in spirit to the probabilisticmodel underlying the RMC. In fact, the two are directly equivalent, a point that was first m... |

65 |
Categorization as probability density estimation
- Ashby, Alfonso-Reese
- 1995
(Show Context)
Citation Context ...at range from strict prototypes (e.g., Reed, 1972) to full exemplar models (e.g., Medin & Schaffer, 1978; Nosofsky, 1986). Recent work has emphasized the “rational” statistical basis of these models (=-=Ashby & Alfonso-Reese, 1995-=-), noting that prototype and exemplar models correspond to different approaches to the “density estimation” problem, in which one infers the probability distribution over stimuli associated with a cat... |

40 |
The Adaptive Character of Thought. Erlbaum
- Anderson
- 1990
(Show Context)
Citation Context ...vidence, RI 02912, USA Daniel J. Navarro (daniel.navarro@adelaide.edu.au) School of Psychology, University of Adelaide, Adelaide SA 5005, Australia Abstract The rational model of categorization (RMC; =-=Anderson, 1990-=-) assumes that categories are learned by clustering similar stimuli together using Bayesian inference. As computing the posterior distribution over all assignments of stimuli to clusters is intractabl... |

21 |
Predictions from uncertain categorizations
- Murphy, Ross
- 1994
(Show Context)
Citation Context ...ore variations on the original model. The rational model of categorization (RMC; Anderson, 1990, 1991) accounts for many of the basic categorization phenomena, although it is not without flaws (e.g., =-=Murphy & Ross, 1994-=-). The RMC uses a flexible representation that can interpolate between prototypes and exemplars by clustering stimuli into groups, 1 adding new clusters to the representation as required. When a new s... |

14 |
Presentation order and recognition of categorically related examples
- Medin, Bettger
- 1994
(Show Context)
Citation Context ...predictions of the local MAP algorithm depend strongly on the presentation order of the stimuli, since cluster assignments are made sequentially and fixed. Order effects are found in human cognition (=-=Medin & Bettger 1994-=-), but are not predicted by the DPMM because of exchangeability. Using data collected by Anderson and Matessa (Anderson, 1990), we explored the strength of the order effects produced by local MAP and ... |

5 | École d’Été de probabilités de Saint-Flour XIII - In - 1985 |