## Perceptual multistability as Markov Chain Monte Carlo inference (2009)

Venue: | Advances in Neural Information Processing Systems 22 |

Citations: | 6 - 3 self |

### BibTeX

@INPROCEEDINGS{Gershman09perceptualmultistability,

author = {Samuel J. Gershman and Edward Vul and Joshua B. Tenenbaum},

title = {Perceptual multistability as Markov Chain Monte Carlo inference},

booktitle = {Advances in Neural Information Processing Systems 22},

year = {2009},

pages = {611--619}

}

### OpenURL

### Abstract

While many perceptual and cognitive phenomena are well described in terms of Bayesian inference, the necessary computations are intractable at the scale of realworld tasks, and it remains unclear how the human mind approximates Bayesian computations algorithmically. We explore the proposal that for some tasks, humans use a form of Markov Chain Monte Carlo to approximate the posterior distribution over hidden variables. As a case study, we show how several phenomena of perceptual multistability can be explained as MCMC inference in simple graphical models for low-level vision. 1

### Citations

3719 | Stochastic Relaxation, Gibbs Distributions and the Bayesian Restoration of Images - Geman, Geman - 1984 |

1157 |
Information Theory, Inference, and Learning Algorithms
- MacKay
- 2003
(Show Context)
Citation Context ...cently proposed that humans approximate complex probabilistic inferences by sampling [19, 14, 21, 6, 4, 24, 23], constructing Monte Carlo estimates similar to those used in Bayesian statistics and AI =-=[16]-=-. A variety of psychological phenomena have natural interpretations in terms of Monte Carlo methods, such as resource limitations [4], stochastic responding [6, 23] and order effects [21, 14]. The Mon... |

286 |
The monte carlo method
- Metropolis, Ulam
- 1949
(Show Context)
Citation Context ...em by drawing samples from a Markov chain that converges to the posterior distribution [16]. There are many variations of MCMC methods but here we will focus on the simplest: the Metropolis algorithm =-=[18]-=-. Each step of the algorithm consists of two stages: a proposal stage and an acceptance stage. An accepted proposal is a sample from a Markov chain that provably converges to the posterior. We will re... |

74 |
Multistable phenomena: changing views in perception
- Leopold, Logothetis
- 1999
(Show Context)
Citation Context ...ence in high-dimensional, sparsely coupled spatial models correspond to several well-known phenomena in visual perception, specifically the dynamics of multistable percepts. Perceptual multistability =-=[13]-=- has long been of interest both phenomenologically and theoretically for models of perception as Bayesian inference [7, 20, 22, 10]. The classic example of perceptual multistability is the Necker cube... |

58 |
The adaptive character of thought. Lawrence Erlbaum
- Anderson
- 1990
(Show Context)
Citation Context ...in simple graphical models for low-level vision. 1 Introduction People appear to make rational statistical inferences from noisy, uncertain input in a wide variety of perceptual and cognitive domains =-=[1, 9]-=-. However, the computations for such inference, even for relatively small problems, are often intractable. For larger problems like those people face in the real world, the space of hypotheses that mu... |

42 | Optimal predictions in everyday cognition
- Griffiths, Tenenbaum
- 2006
(Show Context)
Citation Context ...in simple graphical models for low-level vision. 1 Introduction People appear to make rational statistical inferences from noisy, uncertain input in a wide variety of perceptual and cognitive domains =-=[1, 9]-=-. However, the computations for such inference, even for relatively small problems, are often intractable. For larger problems like those people face in the real world, the space of hypotheses that mu... |

41 | A more rational model of categorization
- Sanborn, Griffiths, et al.
- 2006
(Show Context)
Citation Context ...thmic models for how Bayesian computations are implemented approximately in the human mind. Several authors have recently proposed that humans approximate complex probabilistic inferences by sampling =-=[19, 14, 21, 6, 4, 24, 23]-=-, constructing Monte Carlo estimates similar to those used in Bayesian statistics and AI [16]. A variety of psychological phenomena have natural interpretations in terms of Monte Carlo methods, such a... |

39 |
A Primer on Binocular Rivalry, Including Current Controversies. brain and mind 2
- Blake
- 2001
(Show Context)
Citation Context ...eived to alternate between two different depth configurations (Figure 1A). Another classic phenomenon, extensively studied in psychophysics but less well known outside the field, is binocular rivalry =-=[2]-=-: when incompatible images are presented to the two eyes, subjects report a percept that alternates between the images presented to the left eye and that presented to the right (e.g., Figure 1B). Baye... |

28 | Detecting and predicting changes
- Brown, Steyvers
- 2009
(Show Context)
Citation Context ... close to the Bayesian ideal? Recent work has suggested that people may use approximate inference algorithms similar to those used for solving large-scale problems in Bayesian AI and machine learning =-=[23, 4, 14]-=-. “Rational models” of human cognition at the level of computational theories are often inspired by models for analogous inferences in machine learning. In the same spirit of reverse engineering cogni... |

25 |
Failure of rivalry at low contrast: evidence of a suprathreshold binocular summation process, Vision Res 32
- Liu, Tyler, et al.
- 1992
(Show Context)
Citation Context ...is perceived). This fused perceptual state can be induced most reliably by decreasing the distance in feature space between the two stimuli [11] (Figure 4B) or decreasing the contrast of both stimuli =-=[15]-=-. These relations are shown schematically in Figure 4A. Neither neural, nor algorithmic, nor computational models of rivalry have thus far attempted to explain these findings. In experiments on “fusio... |

23 |
Traveling waves of activity in primary visual cortex during binocular rivalry
- Lee, Blake, et al.
- 2005
(Show Context)
Citation Context ... other eye. One tantalizing feature of these piecemeal percepts is the phenomenon known as traveling waves: subjects tend to perceive a perceptual switch as a “wave” propagating over the visual field =-=[26, 12]-=-: the suppressed stimulus becomes dominant in an isolated location of the visual field and then gradually spreads. These traveling waves reveal an interesting local dynamics during an individual switc... |

22 |
Dynamics of traveling waves in visual perception
- Wilson, Blake, et al.
- 2001
(Show Context)
Citation Context ... other eye. One tantalizing feature of these piecemeal percepts is the phenomenon known as traveling waves: subjects tend to perceive a perceptual switch as a “wave” propagating over the visual field =-=[26, 12]-=-: the suppressed stimulus becomes dominant in an isolated location of the visual field and then gradually spreads. These traveling waves reveal an interesting local dynamics during an individual switc... |

21 | Measuring the crowd within: Probabilistic representations within individuals
- Vul, Pashler
- 2008
(Show Context)
Citation Context ...thmic models for how Bayesian computations are implemented approximately in the human mind. Several authors have recently proposed that humans approximate complex probabilistic inferences by sampling =-=[19, 14, 21, 6, 4, 24, 23]-=-, constructing Monte Carlo estimates similar to those used in Bayesian statistics and AI [16]. A variety of psychological phenomena have natural interpretations in terms of Monte Carlo methods, such a... |

19 | The pigeon as particle filter
- Daw, Courville
- 2008
(Show Context)
Citation Context ...thmic models for how Bayesian computations are implemented approximately in the human mind. Several authors have recently proposed that humans approximate complex probabilistic inferences by sampling =-=[19, 14, 21, 6, 4, 24, 23]-=-, constructing Monte Carlo estimates similar to those used in Bayesian statistics and AI [16]. A variety of psychological phenomena have natural interpretations in terms of Monte Carlo methods, such a... |

18 | A hierarchical model of binocular rivalry
- Dayan
- 1998
(Show Context)
Citation Context ...n, specifically the dynamics of multistable percepts. Perceptual multistability [13] has long been of interest both phenomenologically and theoretically for models of perception as Bayesian inference =-=[7, 20, 22, 10]-=-. The classic example of perceptual multistability is the Necker cube, a 2D line drawing of a cube perceived to alternate between two different depth configurations (Figure 1A). Another classic phenom... |

17 | One and done? Optimal decisions from very few samples - Vul, Goodman, et al. - 2009 |

13 | Performing Bayesian inference with exemplar models - Shi, Feldman, et al. - 2008 |

11 |
Predictive coding explains binocular rivalry: an epistemological review
- Hohwy, Roepstorff, et al.
- 2008
(Show Context)
Citation Context ...n, specifically the dynamics of multistable percepts. Perceptual multistability [13] has long been of interest both phenomenologically and theoretically for models of perception as Bayesian inference =-=[7, 20, 22, 10]-=-. The classic example of perceptual multistability is the Necker cube, a 2D line drawing of a cube perceived to alternate between two different depth configurations (Figure 1A). Another classic phenom... |

10 |
The time course of binocular rivalry reveals a fundamental role of noise
- Brascamp, Ee, et al.
- 2006
(Show Context)
Citation Context ...omenology of bistability in a Necker cube, but it is the phenomenology of binocular rivlary with grating-like stimuli where experiments have shown that substantial time is spent in transition periods =-=[3]-=-. It seems that this is the case in scenarios where a simple planar MRF with nearest neighbor smoothness like the one we’re considering is a good model. To capture the perception of depth in the Necke... |

9 |
Minimal physiological conditions for binocular rivalry and rivalry
- Wilson
- 2007
(Show Context)
Citation Context ...distributions are best [2]. To account for this characteristic behavior, many papers have described neural circuits that could produce switching oscillations with the right stochastic dynamics (e.g., =-=[25]-=-). Existing rational process models of multistability [7, 20, 22] likewise appeal to specific implementational-level constraints to produce 4Figure 2: (A) Simulated timecourse of bistability in the l... |

8 | Temporal dynamics in bistable perception - Mamassian, Goutcher - 2005 |

8 |
Perceptual multistability predicted by search model for Bayesian decisions
- Sundareswara, Schrater
(Show Context)
Citation Context ...n, specifically the dynamics of multistable percepts. Perceptual multistability [13] has long been of interest both phenomenologically and theoretically for models of perception as Bayesian inference =-=[7, 20, 22, 10]-=-. The classic example of perceptual multistability is the Necker cube, a 2D line drawing of a cube perceived to alternate between two different depth configurations (Figure 1A). Another classic phenom... |

5 |
Distance in feature space determines exclusivity in visual rivalry
- Knapen, Kanai, et al.
- 2007
(Show Context)
Citation Context ... superposition of the two stimuli (and hence no alternation is perceived). This fused perceptual state can be induced most reliably by decreasing the distance in feature space between the two stimuli =-=[11]-=- (Figure 4B) or decreasing the contrast of both stimuli [15]. These relations are shown schematically in Figure 4A. Neither neural, nor algorithmic, nor computational models of rivalry have thus far a... |

4 |
Theory and dynamics of perceptual bistability
- Schrater, Sundareswara
- 2006
(Show Context)
Citation Context |

2 |
Contradictory influence of context on predominance during binocular rivalry
- Carter, Campbell, et al.
- 2004
(Show Context)
Citation Context ...(like natural scenes), a more complex and densely interconnected graphical model would be required — in such cases the perceptual switching dynamics will be different. 5Figure 3: (A) Stimuli used by =-=[5]-=- in their experiment. On the top are the standard tilted grating patches presented dichoptically. On the bottom are the tilted grating patches superimposed on a background of rightward-tilting grating... |

1 |
Modeling the effects of memory on human online sentence processing with particle filters
- F, Griffiths
(Show Context)
Citation Context ... close to the Bayesian ideal? Recent work has suggested that people may use approximate inference algorithms similar to those used for solving large-scale problems in Bayesian AI and machine learning =-=[23, 4, 14]-=-. “Rational models” of human cognition at the level of computational theories are often inspired by models for analogous inferences in machine learning. In the same spirit of reverse engineering cogni... |