Limits...
Emergence of functional hierarchy in a multiple timescale neural network model: a humanoid robot experiment.

Yamashita Y, Tani J - PLoS Comput. Biol. (2008)

Bottom Line: The current model neither makes use of separate local modules to represent primitives nor introduces explicit hierarchical structure.In experiments, the proposed network model, coordinating the physical body of a humanoid robot through high-dimensional sensori-motor control, also successfully situated itself within a physical environment.Our results suggest that it is not only the spatial connections between neurons but also the timescales of neural activity that act as important mechanisms leading to functional hierarchy in neural systems.

View Article: PubMed Central - PubMed

Affiliation: Laboratory for Behavior and Dynamic Cognition, RIKEN Brain Science Institute, Wako-shi, Saitama, Japan. yamay@brain.riken.jp

ABSTRACT
It is generally thought that skilled behavior in human beings results from a functional hierarchy of the motor control system, within which reusable motor primitives are flexibly integrated into various sensori-motor sequence patterns. The underlying neural mechanisms governing the way in which continuous sensori-motor flows are segmented into primitives and the way in which series of primitives are integrated into various behavior sequences have, however, not yet been clarified. In earlier studies, this functional hierarchy has been realized through the use of explicit hierarchical structure, with local modules representing motor primitives in the lower level and a higher module representing sequences of primitives switched via additional mechanisms such as gate-selecting. When sequences contain similarities and overlap, however, a conflict arises in such earlier models between generalization and segmentation, induced by this separated modular structure. To address this issue, we propose a different type of neural network model. The current model neither makes use of separate local modules to represent primitives nor introduces explicit hierarchical structure. Rather than forcing architectural hierarchy onto the system, functional hierarchy emerges through a form of self-organization that is based on two distinct types of neurons, each with different time properties ("multiple timescales"). Through the introduction of multiple timescales, continuous sequences of behavior are segmented into reusable primitives, and the primitives, in turn, are flexibly integrated into novel sequences. In experiments, the proposed network model, coordinating the physical body of a humanoid robot through high-dimensional sensori-motor control, also successfully situated itself within a physical environment. Our results suggest that it is not only the spatial connections between neurons but also the timescales of neural activity that act as important mechanisms leading to functional hierarchy in neural systems.

Show MeSH
Primitive representations in fast context units before and afteradditional training.Changes of context activation during each movement before and afteradditional training are visualized in a 2 dimensional space based on theresults of PCA analysis (plotted only for position 3). The four graphson the left side and two graphs on the right side correspond torepresentations before and after additional training, respectively. Thefirst and second movements in the novel sequences learned throughadditional training are colored red (UD and BF) and green (LR andTouch), respectively. The structure of representations corresponding toeach primitive were preserved even after additional training, indicatingthat motor primitives were represented in dynamics of fast contextunits, with novel behavior sequences constructed out of combinations ofthese primitives. UD: up-down, LR: left-right, BF: backward-forward andTouch: touch with single hand behavior.
© Copyright Policy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC2570613&req=5

pcbi-1000220-g008: Primitive representations in fast context units before and afteradditional training.Changes of context activation during each movement before and afteradditional training are visualized in a 2 dimensional space based on theresults of PCA analysis (plotted only for position 3). The four graphson the left side and two graphs on the right side correspond torepresentations before and after additional training, respectively. Thefirst and second movements in the novel sequences learned throughadditional training are colored red (UD and BF) and green (LR andTouch), respectively. The structure of representations corresponding toeach primitive were preserved even after additional training, indicatingthat motor primitives were represented in dynamics of fast contextunits, with novel behavior sequences constructed out of combinations ofthese primitives. UD: up-down, LR: left-right, BF: backward-forward andTouch: touch with single hand behavior.

Mentions: Through training, the robot was able to reproduce perfectly the novel behaviorsequences generalized across object locations, and also managed to successfullyinteract with the physical environment. Figure 7 displays examples of sensori-motorsequences as well as of neural activities of the teaching signal and trainedmodel network interacting with the physical environment. Context unitactivations corresponding to the same behavior were observed to be similar bothin the first basic behavior training and in the additional training. Contextunit activation values corresponding to left-and-right movement in basicbehavior training, for example, were almost identical to context unit activationvalues corresponding to left-and-right movement in the novel sequences used inthe additional training. In order to verify this observation, as in the previoussection, PCA was again conducted for the fast context unit activation valuesduring the execution of novel sequences of behavior. Figure 8 shows examples of changes in thestates of context units for two cases: during the execution of four basicbehavioral patterns following basic pattern training, and during the executionof novel behavior sequences following additional learning.


Emergence of functional hierarchy in a multiple timescale neural network model: a humanoid robot experiment.

Yamashita Y, Tani J - PLoS Comput. Biol. (2008)

Primitive representations in fast context units before and afteradditional training.Changes of context activation during each movement before and afteradditional training are visualized in a 2 dimensional space based on theresults of PCA analysis (plotted only for position 3). The four graphson the left side and two graphs on the right side correspond torepresentations before and after additional training, respectively. Thefirst and second movements in the novel sequences learned throughadditional training are colored red (UD and BF) and green (LR andTouch), respectively. The structure of representations corresponding toeach primitive were preserved even after additional training, indicatingthat motor primitives were represented in dynamics of fast contextunits, with novel behavior sequences constructed out of combinations ofthese primitives. UD: up-down, LR: left-right, BF: backward-forward andTouch: touch with single hand behavior.
© Copyright Policy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC2570613&req=5

pcbi-1000220-g008: Primitive representations in fast context units before and afteradditional training.Changes of context activation during each movement before and afteradditional training are visualized in a 2 dimensional space based on theresults of PCA analysis (plotted only for position 3). The four graphson the left side and two graphs on the right side correspond torepresentations before and after additional training, respectively. Thefirst and second movements in the novel sequences learned throughadditional training are colored red (UD and BF) and green (LR andTouch), respectively. The structure of representations corresponding toeach primitive were preserved even after additional training, indicatingthat motor primitives were represented in dynamics of fast contextunits, with novel behavior sequences constructed out of combinations ofthese primitives. UD: up-down, LR: left-right, BF: backward-forward andTouch: touch with single hand behavior.
Mentions: Through training, the robot was able to reproduce perfectly the novel behaviorsequences generalized across object locations, and also managed to successfullyinteract with the physical environment. Figure 7 displays examples of sensori-motorsequences as well as of neural activities of the teaching signal and trainedmodel network interacting with the physical environment. Context unitactivations corresponding to the same behavior were observed to be similar bothin the first basic behavior training and in the additional training. Contextunit activation values corresponding to left-and-right movement in basicbehavior training, for example, were almost identical to context unit activationvalues corresponding to left-and-right movement in the novel sequences used inthe additional training. In order to verify this observation, as in the previoussection, PCA was again conducted for the fast context unit activation valuesduring the execution of novel sequences of behavior. Figure 8 shows examples of changes in thestates of context units for two cases: during the execution of four basicbehavioral patterns following basic pattern training, and during the executionof novel behavior sequences following additional learning.

Bottom Line: The current model neither makes use of separate local modules to represent primitives nor introduces explicit hierarchical structure.In experiments, the proposed network model, coordinating the physical body of a humanoid robot through high-dimensional sensori-motor control, also successfully situated itself within a physical environment.Our results suggest that it is not only the spatial connections between neurons but also the timescales of neural activity that act as important mechanisms leading to functional hierarchy in neural systems.

View Article: PubMed Central - PubMed

Affiliation: Laboratory for Behavior and Dynamic Cognition, RIKEN Brain Science Institute, Wako-shi, Saitama, Japan. yamay@brain.riken.jp

ABSTRACT
It is generally thought that skilled behavior in human beings results from a functional hierarchy of the motor control system, within which reusable motor primitives are flexibly integrated into various sensori-motor sequence patterns. The underlying neural mechanisms governing the way in which continuous sensori-motor flows are segmented into primitives and the way in which series of primitives are integrated into various behavior sequences have, however, not yet been clarified. In earlier studies, this functional hierarchy has been realized through the use of explicit hierarchical structure, with local modules representing motor primitives in the lower level and a higher module representing sequences of primitives switched via additional mechanisms such as gate-selecting. When sequences contain similarities and overlap, however, a conflict arises in such earlier models between generalization and segmentation, induced by this separated modular structure. To address this issue, we propose a different type of neural network model. The current model neither makes use of separate local modules to represent primitives nor introduces explicit hierarchical structure. Rather than forcing architectural hierarchy onto the system, functional hierarchy emerges through a form of self-organization that is based on two distinct types of neurons, each with different time properties ("multiple timescales"). Through the introduction of multiple timescales, continuous sequences of behavior are segmented into reusable primitives, and the primitives, in turn, are flexibly integrated into novel sequences. In experiments, the proposed network model, coordinating the physical body of a humanoid robot through high-dimensional sensori-motor control, also successfully situated itself within a physical environment. Our results suggest that it is not only the spatial connections between neurons but also the timescales of neural activity that act as important mechanisms leading to functional hierarchy in neural systems.

Show MeSH