Limits...
Bayesian action&perception: representing the world in the brain.

Loeb GE, Fishel JA - Front Neurosci (2014)

Bottom Line: Theories of perception seek to explain how sensory data are processed to identify previously experienced objects, but they usually do not consider the decisions and effort that goes into acquiring the sensory data.In previous studies, a simple robot equipped with a biomimetic tactile sensor and operated according to Bayesian Exploration performed in a manner similar to and actually better than humans on a texture identification task.The biomimetic design of this mechatronic system may provide insights into the neuronal basis of biological action and perception.

View Article: PubMed Central - PubMed

Affiliation: SynTouch LLC Los Angeles, CA, USA ; Department of Biomedical Engineering, University of Southern California Los Angeles, CA, USA.

ABSTRACT
Theories of perception seek to explain how sensory data are processed to identify previously experienced objects, but they usually do not consider the decisions and effort that goes into acquiring the sensory data. Identification of objects according to their tactile properties requires active exploratory movements. The sensory data thereby obtained depend on the details of those movements, which human subjects change rapidly and seemingly capriciously. Bayesian Exploration is an algorithm that uses prior experience to decide which next exploratory movement should provide the most useful data to disambiguate the most likely possibilities. In previous studies, a simple robot equipped with a biomimetic tactile sensor and operated according to Bayesian Exploration performed in a manner similar to and actually better than humans on a texture identification task. Expanding on this, "Bayesian Action&Perception" refers to the construction and querying of an associative memory of previously experienced entities containing both sensory data and the motor programs that elicited them. We hypothesize that this memory can be queried (i) to identify useful next exploratory movements during identification of an unknown entity ("action for perception") or (ii) to characterize whether an unknown entity is fit for purpose ("perception for action") or (iii) to recall what actions might be feasible for a known entity (Gibsonian affordance). The biomimetic design of this mechatronic system may provide insights into the neuronal basis of biological action and perception.

No MeSH data available.


Related in: MedlinePlus

Bayesian representation. The internal representations of entities A and B consist of the previously experienced associations of various motor behaviors with various sensory feedback. The motor repertoire consists of discrete types of movements with continuously adjustable parameters; the sensory dimensions represent processed sensations with continuous ranges of values. Discrimination and identification of entities depends on finding motor-sensory associations that distinguish among the alternatives that are currently most probable. When explored by motor strategy 1, entities A and B result in overlapping sensory percepts a, b, c and g, h, i, respectively. When explored by motor strategy 2, the resulting percepts are non-overlapping for two of the three sensory dimensions (d, e vs. j, l), so this is the better exploratory strategy to pursue. Previous experience has also associated entity B with an affordance consisting of the tendency to produce sensory percepts n, m, o when handled according to motor strategy 3. Once a new object has been identified as entity B (or sufficiently close to it), it is immediately obvious that a behavioral result associated with sensory percepts n, m, and o can be obtained by generating motor strategy 3.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4214374&req=5

Figure 2: Bayesian representation. The internal representations of entities A and B consist of the previously experienced associations of various motor behaviors with various sensory feedback. The motor repertoire consists of discrete types of movements with continuously adjustable parameters; the sensory dimensions represent processed sensations with continuous ranges of values. Discrimination and identification of entities depends on finding motor-sensory associations that distinguish among the alternatives that are currently most probable. When explored by motor strategy 1, entities A and B result in overlapping sensory percepts a, b, c and g, h, i, respectively. When explored by motor strategy 2, the resulting percepts are non-overlapping for two of the three sensory dimensions (d, e vs. j, l), so this is the better exploratory strategy to pursue. Previous experience has also associated entity B with an affordance consisting of the tendency to produce sensory percepts n, m, o when handled according to motor strategy 3. Once a new object has been identified as entity B (or sufficiently close to it), it is immediately obvious that a behavioral result associated with sensory percepts n, m, and o can be obtained by generating motor strategy 3.

Mentions: Bayesian Exploration is an experience-driven algorithm that extends this to decide which next exploratory movement is anticipated to provide the most useful data to disambiguate the current most-probable candidates in an object identification task. It requires a similar experiential database of previously explored objects that associates each object with a set of percepts, each percept consisting of one exploratory movement and a particular sensation thereby obtained (see Figure 2). A graphical representation of this algorithm is provided in Figure 3 (full equations in Fishel and Loeb, 2012a). In summary, the degree of perceptual overlap that would result from making each movement and measuring each sensation for each pair of candidate objects is weighted by the current probabilities that the unknown is one of those two candidate objects. The sum of the weighted overlaps for all object pairs reflects the expected residual ambiguity if that percept were to be measured. The percept that is expected to result in the lowest residual ambiguity wins; its action is performed and the corresponding sensation is compared with expectations to update the Bayesian probabilities.


Bayesian action&perception: representing the world in the brain.

Loeb GE, Fishel JA - Front Neurosci (2014)

Bayesian representation. The internal representations of entities A and B consist of the previously experienced associations of various motor behaviors with various sensory feedback. The motor repertoire consists of discrete types of movements with continuously adjustable parameters; the sensory dimensions represent processed sensations with continuous ranges of values. Discrimination and identification of entities depends on finding motor-sensory associations that distinguish among the alternatives that are currently most probable. When explored by motor strategy 1, entities A and B result in overlapping sensory percepts a, b, c and g, h, i, respectively. When explored by motor strategy 2, the resulting percepts are non-overlapping for two of the three sensory dimensions (d, e vs. j, l), so this is the better exploratory strategy to pursue. Previous experience has also associated entity B with an affordance consisting of the tendency to produce sensory percepts n, m, o when handled according to motor strategy 3. Once a new object has been identified as entity B (or sufficiently close to it), it is immediately obvious that a behavioral result associated with sensory percepts n, m, and o can be obtained by generating motor strategy 3.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4214374&req=5

Figure 2: Bayesian representation. The internal representations of entities A and B consist of the previously experienced associations of various motor behaviors with various sensory feedback. The motor repertoire consists of discrete types of movements with continuously adjustable parameters; the sensory dimensions represent processed sensations with continuous ranges of values. Discrimination and identification of entities depends on finding motor-sensory associations that distinguish among the alternatives that are currently most probable. When explored by motor strategy 1, entities A and B result in overlapping sensory percepts a, b, c and g, h, i, respectively. When explored by motor strategy 2, the resulting percepts are non-overlapping for two of the three sensory dimensions (d, e vs. j, l), so this is the better exploratory strategy to pursue. Previous experience has also associated entity B with an affordance consisting of the tendency to produce sensory percepts n, m, o when handled according to motor strategy 3. Once a new object has been identified as entity B (or sufficiently close to it), it is immediately obvious that a behavioral result associated with sensory percepts n, m, and o can be obtained by generating motor strategy 3.
Mentions: Bayesian Exploration is an experience-driven algorithm that extends this to decide which next exploratory movement is anticipated to provide the most useful data to disambiguate the current most-probable candidates in an object identification task. It requires a similar experiential database of previously explored objects that associates each object with a set of percepts, each percept consisting of one exploratory movement and a particular sensation thereby obtained (see Figure 2). A graphical representation of this algorithm is provided in Figure 3 (full equations in Fishel and Loeb, 2012a). In summary, the degree of perceptual overlap that would result from making each movement and measuring each sensation for each pair of candidate objects is weighted by the current probabilities that the unknown is one of those two candidate objects. The sum of the weighted overlaps for all object pairs reflects the expected residual ambiguity if that percept were to be measured. The percept that is expected to result in the lowest residual ambiguity wins; its action is performed and the corresponding sensation is compared with expectations to update the Bayesian probabilities.

Bottom Line: Theories of perception seek to explain how sensory data are processed to identify previously experienced objects, but they usually do not consider the decisions and effort that goes into acquiring the sensory data.In previous studies, a simple robot equipped with a biomimetic tactile sensor and operated according to Bayesian Exploration performed in a manner similar to and actually better than humans on a texture identification task.The biomimetic design of this mechatronic system may provide insights into the neuronal basis of biological action and perception.

View Article: PubMed Central - PubMed

Affiliation: SynTouch LLC Los Angeles, CA, USA ; Department of Biomedical Engineering, University of Southern California Los Angeles, CA, USA.

ABSTRACT
Theories of perception seek to explain how sensory data are processed to identify previously experienced objects, but they usually do not consider the decisions and effort that goes into acquiring the sensory data. Identification of objects according to their tactile properties requires active exploratory movements. The sensory data thereby obtained depend on the details of those movements, which human subjects change rapidly and seemingly capriciously. Bayesian Exploration is an algorithm that uses prior experience to decide which next exploratory movement should provide the most useful data to disambiguate the most likely possibilities. In previous studies, a simple robot equipped with a biomimetic tactile sensor and operated according to Bayesian Exploration performed in a manner similar to and actually better than humans on a texture identification task. Expanding on this, "Bayesian Action&Perception" refers to the construction and querying of an associative memory of previously experienced entities containing both sensory data and the motor programs that elicited them. We hypothesize that this memory can be queried (i) to identify useful next exploratory movements during identification of an unknown entity ("action for perception") or (ii) to characterize whether an unknown entity is fit for purpose ("perception for action") or (iii) to recall what actions might be feasible for a known entity (Gibsonian affordance). The biomimetic design of this mechatronic system may provide insights into the neuronal basis of biological action and perception.

No MeSH data available.


Related in: MedlinePlus