Disentanglement in conceptual space during sensorimotor interaction
The disentanglement of different objective properties from the external world is the foundation of language development for agents. The basic target of this process is to summarise the common natural properties and then to name it to describe those properties in the future. To realise this purpose,...
Main Authors: | , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
Wiley
2019-10-01
|
Series: | Cognitive Computation and Systems |
Subjects: | |
Online Access: | https://digital-library.theiet.org/content/journals/10.1049/ccs.2019.0007 |
id |
doaj-5ab53ccd21b84e369ceb647d1c4537be |
---|---|
record_format |
Article |
spelling |
doaj-5ab53ccd21b84e369ceb647d1c4537be2021-04-02T16:46:05ZengWileyCognitive Computation and Systems2517-75672019-10-0110.1049/ccs.2019.0007CCS.2019.0007Disentanglement in conceptual space during sensorimotor interactionJunpei Zhong0Tetsuya Ogata1Tetsuya Ogata2Angelo Cangelosi3Angelo Cangelosi4Chenguang Yang5Chenguang Yang6National Institute of Advanced Science and Technology (AIST)National Institute of Advanced Science and Technology (AIST)National Institute of Advanced Science and Technology (AIST)University of ManchesterUniversity of ManchesterUniversity of the West of EnglandUniversity of the West of EnglandThe disentanglement of different objective properties from the external world is the foundation of language development for agents. The basic target of this process is to summarise the common natural properties and then to name it to describe those properties in the future. To realise this purpose, a new learning model is introduced for the disentanglement of several sensorimotor concepts (e.g. sizes, colours and shapes of objects) while the causal relationship is being learnt during interaction without much a priori experience and external instructions. This learning model links predictive deep neural models and the variational auto-encoder (VAE) and provides the possibility that the independent concepts can be extracted and disentangled from both perception and action. Moreover, such extraction is further learnt by VAE to memorise their common statistical features. The authors examine this model in the affordance learning setting, where the robot is trying to learn to disentangle about shapes of the tools and objects. The results show that such a process can be found in the neural activities of the [inline-formula]-VAE unit, which indicate that using similar VAE models is a promising way to learn the concepts, and thereby to learn the causal relationship of the sensorimotor interaction.https://digital-library.theiet.org/content/journals/10.1049/ccs.2019.0007neural netshuman-robot interactionlearning (artificial intelligence)multi-agent systemsmulti-robot systemsstatistical analysissensorimotor interactiondisentanglementconceptual spaceintelligent agentssensorimotor conceptslearning modeldeep neural modelsvariational auto-encoderaffordance learning settingvaerobot |
collection |
DOAJ |
language |
English |
format |
Article |
sources |
DOAJ |
author |
Junpei Zhong Tetsuya Ogata Tetsuya Ogata Angelo Cangelosi Angelo Cangelosi Chenguang Yang Chenguang Yang |
spellingShingle |
Junpei Zhong Tetsuya Ogata Tetsuya Ogata Angelo Cangelosi Angelo Cangelosi Chenguang Yang Chenguang Yang Disentanglement in conceptual space during sensorimotor interaction Cognitive Computation and Systems neural nets human-robot interaction learning (artificial intelligence) multi-agent systems multi-robot systems statistical analysis sensorimotor interaction disentanglement conceptual space intelligent agents sensorimotor concepts learning model deep neural models variational auto-encoder affordance learning setting vae robot |
author_facet |
Junpei Zhong Tetsuya Ogata Tetsuya Ogata Angelo Cangelosi Angelo Cangelosi Chenguang Yang Chenguang Yang |
author_sort |
Junpei Zhong |
title |
Disentanglement in conceptual space during sensorimotor interaction |
title_short |
Disentanglement in conceptual space during sensorimotor interaction |
title_full |
Disentanglement in conceptual space during sensorimotor interaction |
title_fullStr |
Disentanglement in conceptual space during sensorimotor interaction |
title_full_unstemmed |
Disentanglement in conceptual space during sensorimotor interaction |
title_sort |
disentanglement in conceptual space during sensorimotor interaction |
publisher |
Wiley |
series |
Cognitive Computation and Systems |
issn |
2517-7567 |
publishDate |
2019-10-01 |
description |
The disentanglement of different objective properties from the external world is the foundation of language development for agents. The basic target of this process is to summarise the common natural properties and then to name it to describe those properties in the future. To realise this purpose, a new learning model is introduced for the disentanglement of several sensorimotor concepts (e.g. sizes, colours and shapes of objects) while the causal relationship is being learnt during interaction without much a priori experience and external instructions. This learning model links predictive deep neural models and the variational auto-encoder (VAE) and provides the possibility that the independent concepts can be extracted and disentangled from both perception and action. Moreover, such extraction is further learnt by VAE to memorise their common statistical features. The authors examine this model in the affordance learning setting, where the robot is trying to learn to disentangle about shapes of the tools and objects. The results show that such a process can be found in the neural activities of the [inline-formula]-VAE unit, which indicate that using similar VAE models is a promising way to learn the concepts, and thereby to learn the causal relationship of the sensorimotor interaction. |
topic |
neural nets human-robot interaction learning (artificial intelligence) multi-agent systems multi-robot systems statistical analysis sensorimotor interaction disentanglement conceptual space intelligent agents sensorimotor concepts learning model deep neural models variational auto-encoder affordance learning setting vae robot |
url |
https://digital-library.theiet.org/content/journals/10.1049/ccs.2019.0007 |
work_keys_str_mv |
AT junpeizhong disentanglementinconceptualspaceduringsensorimotorinteraction AT tetsuyaogata disentanglementinconceptualspaceduringsensorimotorinteraction AT tetsuyaogata disentanglementinconceptualspaceduringsensorimotorinteraction AT angelocangelosi disentanglementinconceptualspaceduringsensorimotorinteraction AT angelocangelosi disentanglementinconceptualspaceduringsensorimotorinteraction AT chenguangyang disentanglementinconceptualspaceduringsensorimotorinteraction AT chenguangyang disentanglementinconceptualspaceduringsensorimotorinteraction |
_version_ |
1721555421230006272 |