Behavior Generation with Latent Actions
Generative modeling of complex behaviors from labeled datasets has been a longstanding problem in decision making. Unlike language or image generation, decision making requires modeling actions - continuous-valued vectors that are multimodal in their distribution, potentially drawn from uncurated so...
Gespeichert in:
Veröffentlicht in: | arXiv.org 2024-06 |
---|---|
Hauptverfasser: | , , , , , |
Format: | Artikel |
Sprache: | eng |
Schlagworte: | |
Online-Zugang: | Volltext |
Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
container_end_page | |
---|---|
container_issue | |
container_start_page | |
container_title | arXiv.org |
container_volume | |
creator | Lee, Seungjae Wang, Yibin Etukuru, Haritheja H Jin Kim Nur Muhammad Mahi Shafiullah Pinto, Lerrel |
description | Generative modeling of complex behaviors from labeled datasets has been a longstanding problem in decision making. Unlike language or image generation, decision making requires modeling actions - continuous-valued vectors that are multimodal in their distribution, potentially drawn from uncurated sources, where generation errors can compound in sequential prediction. A recent class of models called Behavior Transformers (BeT) addresses this by discretizing actions using k-means clustering to capture different modes. However, k-means struggles to scale for high-dimensional action spaces or long sequences, and lacks gradient information, and thus BeT suffers in modeling long-range actions. In this work, we present Vector-Quantized Behavior Transformer (VQ-BeT), a versatile model for behavior generation that handles multimodal action prediction, conditional generation, and partial observations. VQ-BeT augments BeT by tokenizing continuous actions with a hierarchical vector quantization module. Across seven environments including simulated manipulation, autonomous driving, and robotics, VQ-BeT improves on state-of-the-art models such as BeT and Diffusion Policies. Importantly, we demonstrate VQ-BeT's improved ability to capture behavior modes while accelerating inference speed 5x over Diffusion Policies. Videos and code can be found https://sjlee.cc/vq-bet |
format | Article |
fullrecord | <record><control><sourceid>proquest</sourceid><recordid>TN_cdi_proquest_journals_2938060726</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2938060726</sourcerecordid><originalsourceid>FETCH-proquest_journals_29380607263</originalsourceid><addsrcrecordid>eNpjYuA0MjY21LUwMTLiYOAtLs4yMDAwMjM3MjU15mRQd0rNSCzLzC9ScE_NSy1KLMnMz1MozyzJUPBJLEnNK1FwTAYJFfMwsKYl5hSn8kJpbgZlN9cQZw_dgqL8wtLU4pL4rPzSojygVLyRpbGFgZmBuZGZMXGqAGmdL7M</addsrcrecordid><sourcetype>Aggregation Database</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype><pqid>2938060726</pqid></control><display><type>article</type><title>Behavior Generation with Latent Actions</title><source>Free E- Journals</source><creator>Lee, Seungjae ; Wang, Yibin ; Etukuru, Haritheja ; H Jin Kim ; Nur Muhammad Mahi Shafiullah ; Pinto, Lerrel</creator><creatorcontrib>Lee, Seungjae ; Wang, Yibin ; Etukuru, Haritheja ; H Jin Kim ; Nur Muhammad Mahi Shafiullah ; Pinto, Lerrel</creatorcontrib><description>Generative modeling of complex behaviors from labeled datasets has been a longstanding problem in decision making. Unlike language or image generation, decision making requires modeling actions - continuous-valued vectors that are multimodal in their distribution, potentially drawn from uncurated sources, where generation errors can compound in sequential prediction. A recent class of models called Behavior Transformers (BeT) addresses this by discretizing actions using k-means clustering to capture different modes. However, k-means struggles to scale for high-dimensional action spaces or long sequences, and lacks gradient information, and thus BeT suffers in modeling long-range actions. In this work, we present Vector-Quantized Behavior Transformer (VQ-BeT), a versatile model for behavior generation that handles multimodal action prediction, conditional generation, and partial observations. VQ-BeT augments BeT by tokenizing continuous actions with a hierarchical vector quantization module. Across seven environments including simulated manipulation, autonomous driving, and robotics, VQ-BeT improves on state-of-the-art models such as BeT and Diffusion Policies. Importantly, we demonstrate VQ-BeT's improved ability to capture behavior modes while accelerating inference speed 5x over Diffusion Policies. Videos and code can be found https://sjlee.cc/vq-bet</description><identifier>EISSN: 2331-8422</identifier><language>eng</language><publisher>Ithaca: Cornell University Library, arXiv.org</publisher><subject>Cluster analysis ; Clustering ; Decision making ; Diffusion rate ; Image processing ; Modelling ; Policies ; Robotics ; Transformers ; Vector quantization</subject><ispartof>arXiv.org, 2024-06</ispartof><rights>2024. This work is published under http://creativecommons.org/licenses/by/4.0/ (the “License”). Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License.</rights><oa>free_for_read</oa><woscitedreferencessubscribed>false</woscitedreferencessubscribed></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><link.rule.ids>780,784</link.rule.ids></links><search><creatorcontrib>Lee, Seungjae</creatorcontrib><creatorcontrib>Wang, Yibin</creatorcontrib><creatorcontrib>Etukuru, Haritheja</creatorcontrib><creatorcontrib>H Jin Kim</creatorcontrib><creatorcontrib>Nur Muhammad Mahi Shafiullah</creatorcontrib><creatorcontrib>Pinto, Lerrel</creatorcontrib><title>Behavior Generation with Latent Actions</title><title>arXiv.org</title><description>Generative modeling of complex behaviors from labeled datasets has been a longstanding problem in decision making. Unlike language or image generation, decision making requires modeling actions - continuous-valued vectors that are multimodal in their distribution, potentially drawn from uncurated sources, where generation errors can compound in sequential prediction. A recent class of models called Behavior Transformers (BeT) addresses this by discretizing actions using k-means clustering to capture different modes. However, k-means struggles to scale for high-dimensional action spaces or long sequences, and lacks gradient information, and thus BeT suffers in modeling long-range actions. In this work, we present Vector-Quantized Behavior Transformer (VQ-BeT), a versatile model for behavior generation that handles multimodal action prediction, conditional generation, and partial observations. VQ-BeT augments BeT by tokenizing continuous actions with a hierarchical vector quantization module. Across seven environments including simulated manipulation, autonomous driving, and robotics, VQ-BeT improves on state-of-the-art models such as BeT and Diffusion Policies. Importantly, we demonstrate VQ-BeT's improved ability to capture behavior modes while accelerating inference speed 5x over Diffusion Policies. Videos and code can be found https://sjlee.cc/vq-bet</description><subject>Cluster analysis</subject><subject>Clustering</subject><subject>Decision making</subject><subject>Diffusion rate</subject><subject>Image processing</subject><subject>Modelling</subject><subject>Policies</subject><subject>Robotics</subject><subject>Transformers</subject><subject>Vector quantization</subject><issn>2331-8422</issn><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2024</creationdate><recordtype>article</recordtype><sourceid>ABUWG</sourceid><sourceid>AFKRA</sourceid><sourceid>AZQEC</sourceid><sourceid>BENPR</sourceid><sourceid>CCPQU</sourceid><sourceid>DWQXO</sourceid><recordid>eNpjYuA0MjY21LUwMTLiYOAtLs4yMDAwMjM3MjU15mRQd0rNSCzLzC9ScE_NSy1KLMnMz1MozyzJUPBJLEnNK1FwTAYJFfMwsKYl5hSn8kJpbgZlN9cQZw_dgqL8wtLU4pL4rPzSojygVLyRpbGFgZmBuZGZMXGqAGmdL7M</recordid><startdate>20240628</startdate><enddate>20240628</enddate><creator>Lee, Seungjae</creator><creator>Wang, Yibin</creator><creator>Etukuru, Haritheja</creator><creator>H Jin Kim</creator><creator>Nur Muhammad Mahi Shafiullah</creator><creator>Pinto, Lerrel</creator><general>Cornell University Library, arXiv.org</general><scope>8FE</scope><scope>8FG</scope><scope>ABJCF</scope><scope>ABUWG</scope><scope>AFKRA</scope><scope>AZQEC</scope><scope>BENPR</scope><scope>BGLVJ</scope><scope>CCPQU</scope><scope>DWQXO</scope><scope>HCIFZ</scope><scope>L6V</scope><scope>M7S</scope><scope>PIMPY</scope><scope>PQEST</scope><scope>PQQKQ</scope><scope>PQUKI</scope><scope>PRINS</scope><scope>PTHSS</scope></search><sort><creationdate>20240628</creationdate><title>Behavior Generation with Latent Actions</title><author>Lee, Seungjae ; Wang, Yibin ; Etukuru, Haritheja ; H Jin Kim ; Nur Muhammad Mahi Shafiullah ; Pinto, Lerrel</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-proquest_journals_29380607263</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2024</creationdate><topic>Cluster analysis</topic><topic>Clustering</topic><topic>Decision making</topic><topic>Diffusion rate</topic><topic>Image processing</topic><topic>Modelling</topic><topic>Policies</topic><topic>Robotics</topic><topic>Transformers</topic><topic>Vector quantization</topic><toplevel>online_resources</toplevel><creatorcontrib>Lee, Seungjae</creatorcontrib><creatorcontrib>Wang, Yibin</creatorcontrib><creatorcontrib>Etukuru, Haritheja</creatorcontrib><creatorcontrib>H Jin Kim</creatorcontrib><creatorcontrib>Nur Muhammad Mahi Shafiullah</creatorcontrib><creatorcontrib>Pinto, Lerrel</creatorcontrib><collection>ProQuest SciTech Collection</collection><collection>ProQuest Technology Collection</collection><collection>Materials Science & Engineering Collection</collection><collection>ProQuest Central (Alumni Edition)</collection><collection>ProQuest Central UK/Ireland</collection><collection>ProQuest Central Essentials</collection><collection>ProQuest Central</collection><collection>Technology Collection</collection><collection>ProQuest One Community College</collection><collection>ProQuest Central Korea</collection><collection>SciTech Premium Collection</collection><collection>ProQuest Engineering Collection</collection><collection>Engineering Database</collection><collection>Publicly Available Content Database</collection><collection>ProQuest One Academic Eastern Edition (DO NOT USE)</collection><collection>ProQuest One Academic</collection><collection>ProQuest One Academic UKI Edition</collection><collection>ProQuest Central China</collection><collection>Engineering Collection</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext</fulltext></delivery><addata><au>Lee, Seungjae</au><au>Wang, Yibin</au><au>Etukuru, Haritheja</au><au>H Jin Kim</au><au>Nur Muhammad Mahi Shafiullah</au><au>Pinto, Lerrel</au><format>book</format><genre>document</genre><ristype>GEN</ristype><atitle>Behavior Generation with Latent Actions</atitle><jtitle>arXiv.org</jtitle><date>2024-06-28</date><risdate>2024</risdate><eissn>2331-8422</eissn><abstract>Generative modeling of complex behaviors from labeled datasets has been a longstanding problem in decision making. Unlike language or image generation, decision making requires modeling actions - continuous-valued vectors that are multimodal in their distribution, potentially drawn from uncurated sources, where generation errors can compound in sequential prediction. A recent class of models called Behavior Transformers (BeT) addresses this by discretizing actions using k-means clustering to capture different modes. However, k-means struggles to scale for high-dimensional action spaces or long sequences, and lacks gradient information, and thus BeT suffers in modeling long-range actions. In this work, we present Vector-Quantized Behavior Transformer (VQ-BeT), a versatile model for behavior generation that handles multimodal action prediction, conditional generation, and partial observations. VQ-BeT augments BeT by tokenizing continuous actions with a hierarchical vector quantization module. Across seven environments including simulated manipulation, autonomous driving, and robotics, VQ-BeT improves on state-of-the-art models such as BeT and Diffusion Policies. Importantly, we demonstrate VQ-BeT's improved ability to capture behavior modes while accelerating inference speed 5x over Diffusion Policies. Videos and code can be found https://sjlee.cc/vq-bet</abstract><cop>Ithaca</cop><pub>Cornell University Library, arXiv.org</pub><oa>free_for_read</oa></addata></record> |
fulltext | fulltext |
identifier | EISSN: 2331-8422 |
ispartof | arXiv.org, 2024-06 |
issn | 2331-8422 |
language | eng |
recordid | cdi_proquest_journals_2938060726 |
source | Free E- Journals |
subjects | Cluster analysis Clustering Decision making Diffusion rate Image processing Modelling Policies Robotics Transformers Vector quantization |
title | Behavior Generation with Latent Actions |
url | https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2024-12-25T16%3A11%3A53IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-proquest&rft_val_fmt=info:ofi/fmt:kev:mtx:book&rft.genre=document&rft.atitle=Behavior%20Generation%20with%20Latent%20Actions&rft.jtitle=arXiv.org&rft.au=Lee,%20Seungjae&rft.date=2024-06-28&rft.eissn=2331-8422&rft_id=info:doi/&rft_dat=%3Cproquest%3E2938060726%3C/proquest%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_pqid=2938060726&rft_id=info:pmid/&rfr_iscdi=true |