Deep imitation learning for 3D navigation tasks
Abstract Deep learning techniques have shown success in learning from raw high-dimensional data in various applications. While deep reinforcement learning is recently gaining popularity as a method to train intelligent agents, utilizing deep learning in imitation learning has been scarcely explored....
Ausführliche Beschreibung
Autor*in: |
Hussein, Ahmed [verfasserIn] |
---|
Format: |
Artikel |
---|---|
Sprache: |
Englisch |
Erschienen: |
2017 |
---|
Schlagwörter: |
---|
Anmerkung: |
© The Author(s) 2018 |
---|
Übergeordnetes Werk: |
Enthalten in: Neural computing & applications - Springer London, 1993, 29(2017), 7 vom: 04. Dez., Seite 389-404 |
---|---|
Übergeordnetes Werk: |
volume:29 ; year:2017 ; number:7 ; day:04 ; month:12 ; pages:389-404 |
Links: |
---|
DOI / URN: |
10.1007/s00521-017-3241-z |
---|
Katalog-ID: |
OLC2025604963 |
---|
LEADER | 01000caa a22002652 4500 | ||
---|---|---|---|
001 | OLC2025604963 | ||
003 | DE-627 | ||
005 | 20230502114736.0 | ||
007 | tu | ||
008 | 200820s2017 xx ||||| 00| ||eng c | ||
024 | 7 | |a 10.1007/s00521-017-3241-z |2 doi | |
035 | |a (DE-627)OLC2025604963 | ||
035 | |a (DE-He213)s00521-017-3241-z-p | ||
040 | |a DE-627 |b ger |c DE-627 |e rakwb | ||
041 | |a eng | ||
082 | 0 | 4 | |a 004 |q VZ |
100 | 1 | |a Hussein, Ahmed |e verfasserin |0 (orcid)0000-0001-5227-9929 |4 aut | |
245 | 1 | 0 | |a Deep imitation learning for 3D navigation tasks |
264 | 1 | |c 2017 | |
336 | |a Text |b txt |2 rdacontent | ||
337 | |a ohne Hilfsmittel zu benutzen |b n |2 rdamedia | ||
338 | |a Band |b nc |2 rdacarrier | ||
500 | |a © The Author(s) 2018 | ||
520 | |a Abstract Deep learning techniques have shown success in learning from raw high-dimensional data in various applications. While deep reinforcement learning is recently gaining popularity as a method to train intelligent agents, utilizing deep learning in imitation learning has been scarcely explored. Imitation learning can be an efficient method to teach intelligent agents by providing a set of demonstrations to learn from. However, generalizing to situations that are not represented in the demonstrations can be challenging, especially in 3D environments. In this paper, we propose a deep imitation learning method to learn navigation tasks from demonstrations in a 3D environment. The supervised policy is refined using active learning in order to generalize to unseen situations. This approach is compared to two popular deep reinforcement learning techniques: deep-Q-networks and Asynchronous actor-critic (A3C). The proposed method as well as the reinforcement learning methods employ deep convolutional neural networks and learn directly from raw visual input. Methods for combining learning from demonstrations and experience are also investigated. This combination aims to join the generalization ability of learning by experience with the efficiency of learning by imitation. The proposed methods are evaluated on 4 navigation tasks in a 3D simulated environment. Navigation tasks are a typical problem that is relevant to many real applications. They pose the challenge of requiring demonstrations of long trajectories to reach the target and only providing delayed rewards (usually terminal) to the agent. The experiments show that the proposed method can successfully learn navigation tasks from raw visual input while learning from experience methods fail to learn an effective policy. Moreover, it is shown that active learning can significantly improve the performance of the initially learned policy using a small number of active samples. | ||
650 | 4 | |a Deep learning | |
650 | 4 | |a Convolutional neural networks | |
650 | 4 | |a Learning from demonstrations | |
650 | 4 | |a Reinforcement learning | |
650 | 4 | |a Active learning | |
650 | 4 | |a 3D navigation | |
650 | 4 | |a Benchmarking | |
700 | 1 | |a Elyan, Eyad |4 aut | |
700 | 1 | |a Gaber, Mohamed Medhat |4 aut | |
700 | 1 | |a Jayne, Chrisina |4 aut | |
773 | 0 | 8 | |i Enthalten in |t Neural computing & applications |d Springer London, 1993 |g 29(2017), 7 vom: 04. Dez., Seite 389-404 |w (DE-627)165669608 |w (DE-600)1136944-9 |w (DE-576)032873050 |x 0941-0643 |7 nnns |
773 | 1 | 8 | |g volume:29 |g year:2017 |g number:7 |g day:04 |g month:12 |g pages:389-404 |
856 | 4 | 1 | |u https://doi.org/10.1007/s00521-017-3241-z |z lizenzpflichtig |3 Volltext |
912 | |a GBV_USEFLAG_A | ||
912 | |a SYSFLAG_A | ||
912 | |a GBV_OLC | ||
912 | |a SSG-OLC-MAT | ||
912 | |a GBV_ILN_70 | ||
912 | |a GBV_ILN_2018 | ||
912 | |a GBV_ILN_4046 | ||
912 | |a GBV_ILN_4277 | ||
951 | |a AR | ||
952 | |d 29 |j 2017 |e 7 |b 04 |c 12 |h 389-404 |
author_variant |
a h ah e e ee m m g mm mmg c j cj |
---|---|
matchkey_str |
article:09410643:2017----::epmttolannfrdai |
hierarchy_sort_str |
2017 |
publishDate |
2017 |
allfields |
10.1007/s00521-017-3241-z doi (DE-627)OLC2025604963 (DE-He213)s00521-017-3241-z-p DE-627 ger DE-627 rakwb eng 004 VZ Hussein, Ahmed verfasserin (orcid)0000-0001-5227-9929 aut Deep imitation learning for 3D navigation tasks 2017 Text txt rdacontent ohne Hilfsmittel zu benutzen n rdamedia Band nc rdacarrier © The Author(s) 2018 Abstract Deep learning techniques have shown success in learning from raw high-dimensional data in various applications. While deep reinforcement learning is recently gaining popularity as a method to train intelligent agents, utilizing deep learning in imitation learning has been scarcely explored. Imitation learning can be an efficient method to teach intelligent agents by providing a set of demonstrations to learn from. However, generalizing to situations that are not represented in the demonstrations can be challenging, especially in 3D environments. In this paper, we propose a deep imitation learning method to learn navigation tasks from demonstrations in a 3D environment. The supervised policy is refined using active learning in order to generalize to unseen situations. This approach is compared to two popular deep reinforcement learning techniques: deep-Q-networks and Asynchronous actor-critic (A3C). The proposed method as well as the reinforcement learning methods employ deep convolutional neural networks and learn directly from raw visual input. Methods for combining learning from demonstrations and experience are also investigated. This combination aims to join the generalization ability of learning by experience with the efficiency of learning by imitation. The proposed methods are evaluated on 4 navigation tasks in a 3D simulated environment. Navigation tasks are a typical problem that is relevant to many real applications. They pose the challenge of requiring demonstrations of long trajectories to reach the target and only providing delayed rewards (usually terminal) to the agent. The experiments show that the proposed method can successfully learn navigation tasks from raw visual input while learning from experience methods fail to learn an effective policy. Moreover, it is shown that active learning can significantly improve the performance of the initially learned policy using a small number of active samples. Deep learning Convolutional neural networks Learning from demonstrations Reinforcement learning Active learning 3D navigation Benchmarking Elyan, Eyad aut Gaber, Mohamed Medhat aut Jayne, Chrisina aut Enthalten in Neural computing & applications Springer London, 1993 29(2017), 7 vom: 04. Dez., Seite 389-404 (DE-627)165669608 (DE-600)1136944-9 (DE-576)032873050 0941-0643 nnns volume:29 year:2017 number:7 day:04 month:12 pages:389-404 https://doi.org/10.1007/s00521-017-3241-z lizenzpflichtig Volltext GBV_USEFLAG_A SYSFLAG_A GBV_OLC SSG-OLC-MAT GBV_ILN_70 GBV_ILN_2018 GBV_ILN_4046 GBV_ILN_4277 AR 29 2017 7 04 12 389-404 |
spelling |
10.1007/s00521-017-3241-z doi (DE-627)OLC2025604963 (DE-He213)s00521-017-3241-z-p DE-627 ger DE-627 rakwb eng 004 VZ Hussein, Ahmed verfasserin (orcid)0000-0001-5227-9929 aut Deep imitation learning for 3D navigation tasks 2017 Text txt rdacontent ohne Hilfsmittel zu benutzen n rdamedia Band nc rdacarrier © The Author(s) 2018 Abstract Deep learning techniques have shown success in learning from raw high-dimensional data in various applications. While deep reinforcement learning is recently gaining popularity as a method to train intelligent agents, utilizing deep learning in imitation learning has been scarcely explored. Imitation learning can be an efficient method to teach intelligent agents by providing a set of demonstrations to learn from. However, generalizing to situations that are not represented in the demonstrations can be challenging, especially in 3D environments. In this paper, we propose a deep imitation learning method to learn navigation tasks from demonstrations in a 3D environment. The supervised policy is refined using active learning in order to generalize to unseen situations. This approach is compared to two popular deep reinforcement learning techniques: deep-Q-networks and Asynchronous actor-critic (A3C). The proposed method as well as the reinforcement learning methods employ deep convolutional neural networks and learn directly from raw visual input. Methods for combining learning from demonstrations and experience are also investigated. This combination aims to join the generalization ability of learning by experience with the efficiency of learning by imitation. The proposed methods are evaluated on 4 navigation tasks in a 3D simulated environment. Navigation tasks are a typical problem that is relevant to many real applications. They pose the challenge of requiring demonstrations of long trajectories to reach the target and only providing delayed rewards (usually terminal) to the agent. The experiments show that the proposed method can successfully learn navigation tasks from raw visual input while learning from experience methods fail to learn an effective policy. Moreover, it is shown that active learning can significantly improve the performance of the initially learned policy using a small number of active samples. Deep learning Convolutional neural networks Learning from demonstrations Reinforcement learning Active learning 3D navigation Benchmarking Elyan, Eyad aut Gaber, Mohamed Medhat aut Jayne, Chrisina aut Enthalten in Neural computing & applications Springer London, 1993 29(2017), 7 vom: 04. Dez., Seite 389-404 (DE-627)165669608 (DE-600)1136944-9 (DE-576)032873050 0941-0643 nnns volume:29 year:2017 number:7 day:04 month:12 pages:389-404 https://doi.org/10.1007/s00521-017-3241-z lizenzpflichtig Volltext GBV_USEFLAG_A SYSFLAG_A GBV_OLC SSG-OLC-MAT GBV_ILN_70 GBV_ILN_2018 GBV_ILN_4046 GBV_ILN_4277 AR 29 2017 7 04 12 389-404 |
allfields_unstemmed |
10.1007/s00521-017-3241-z doi (DE-627)OLC2025604963 (DE-He213)s00521-017-3241-z-p DE-627 ger DE-627 rakwb eng 004 VZ Hussein, Ahmed verfasserin (orcid)0000-0001-5227-9929 aut Deep imitation learning for 3D navigation tasks 2017 Text txt rdacontent ohne Hilfsmittel zu benutzen n rdamedia Band nc rdacarrier © The Author(s) 2018 Abstract Deep learning techniques have shown success in learning from raw high-dimensional data in various applications. While deep reinforcement learning is recently gaining popularity as a method to train intelligent agents, utilizing deep learning in imitation learning has been scarcely explored. Imitation learning can be an efficient method to teach intelligent agents by providing a set of demonstrations to learn from. However, generalizing to situations that are not represented in the demonstrations can be challenging, especially in 3D environments. In this paper, we propose a deep imitation learning method to learn navigation tasks from demonstrations in a 3D environment. The supervised policy is refined using active learning in order to generalize to unseen situations. This approach is compared to two popular deep reinforcement learning techniques: deep-Q-networks and Asynchronous actor-critic (A3C). The proposed method as well as the reinforcement learning methods employ deep convolutional neural networks and learn directly from raw visual input. Methods for combining learning from demonstrations and experience are also investigated. This combination aims to join the generalization ability of learning by experience with the efficiency of learning by imitation. The proposed methods are evaluated on 4 navigation tasks in a 3D simulated environment. Navigation tasks are a typical problem that is relevant to many real applications. They pose the challenge of requiring demonstrations of long trajectories to reach the target and only providing delayed rewards (usually terminal) to the agent. The experiments show that the proposed method can successfully learn navigation tasks from raw visual input while learning from experience methods fail to learn an effective policy. Moreover, it is shown that active learning can significantly improve the performance of the initially learned policy using a small number of active samples. Deep learning Convolutional neural networks Learning from demonstrations Reinforcement learning Active learning 3D navigation Benchmarking Elyan, Eyad aut Gaber, Mohamed Medhat aut Jayne, Chrisina aut Enthalten in Neural computing & applications Springer London, 1993 29(2017), 7 vom: 04. Dez., Seite 389-404 (DE-627)165669608 (DE-600)1136944-9 (DE-576)032873050 0941-0643 nnns volume:29 year:2017 number:7 day:04 month:12 pages:389-404 https://doi.org/10.1007/s00521-017-3241-z lizenzpflichtig Volltext GBV_USEFLAG_A SYSFLAG_A GBV_OLC SSG-OLC-MAT GBV_ILN_70 GBV_ILN_2018 GBV_ILN_4046 GBV_ILN_4277 AR 29 2017 7 04 12 389-404 |
allfieldsGer |
10.1007/s00521-017-3241-z doi (DE-627)OLC2025604963 (DE-He213)s00521-017-3241-z-p DE-627 ger DE-627 rakwb eng 004 VZ Hussein, Ahmed verfasserin (orcid)0000-0001-5227-9929 aut Deep imitation learning for 3D navigation tasks 2017 Text txt rdacontent ohne Hilfsmittel zu benutzen n rdamedia Band nc rdacarrier © The Author(s) 2018 Abstract Deep learning techniques have shown success in learning from raw high-dimensional data in various applications. While deep reinforcement learning is recently gaining popularity as a method to train intelligent agents, utilizing deep learning in imitation learning has been scarcely explored. Imitation learning can be an efficient method to teach intelligent agents by providing a set of demonstrations to learn from. However, generalizing to situations that are not represented in the demonstrations can be challenging, especially in 3D environments. In this paper, we propose a deep imitation learning method to learn navigation tasks from demonstrations in a 3D environment. The supervised policy is refined using active learning in order to generalize to unseen situations. This approach is compared to two popular deep reinforcement learning techniques: deep-Q-networks and Asynchronous actor-critic (A3C). The proposed method as well as the reinforcement learning methods employ deep convolutional neural networks and learn directly from raw visual input. Methods for combining learning from demonstrations and experience are also investigated. This combination aims to join the generalization ability of learning by experience with the efficiency of learning by imitation. The proposed methods are evaluated on 4 navigation tasks in a 3D simulated environment. Navigation tasks are a typical problem that is relevant to many real applications. They pose the challenge of requiring demonstrations of long trajectories to reach the target and only providing delayed rewards (usually terminal) to the agent. The experiments show that the proposed method can successfully learn navigation tasks from raw visual input while learning from experience methods fail to learn an effective policy. Moreover, it is shown that active learning can significantly improve the performance of the initially learned policy using a small number of active samples. Deep learning Convolutional neural networks Learning from demonstrations Reinforcement learning Active learning 3D navigation Benchmarking Elyan, Eyad aut Gaber, Mohamed Medhat aut Jayne, Chrisina aut Enthalten in Neural computing & applications Springer London, 1993 29(2017), 7 vom: 04. Dez., Seite 389-404 (DE-627)165669608 (DE-600)1136944-9 (DE-576)032873050 0941-0643 nnns volume:29 year:2017 number:7 day:04 month:12 pages:389-404 https://doi.org/10.1007/s00521-017-3241-z lizenzpflichtig Volltext GBV_USEFLAG_A SYSFLAG_A GBV_OLC SSG-OLC-MAT GBV_ILN_70 GBV_ILN_2018 GBV_ILN_4046 GBV_ILN_4277 AR 29 2017 7 04 12 389-404 |
allfieldsSound |
10.1007/s00521-017-3241-z doi (DE-627)OLC2025604963 (DE-He213)s00521-017-3241-z-p DE-627 ger DE-627 rakwb eng 004 VZ Hussein, Ahmed verfasserin (orcid)0000-0001-5227-9929 aut Deep imitation learning for 3D navigation tasks 2017 Text txt rdacontent ohne Hilfsmittel zu benutzen n rdamedia Band nc rdacarrier © The Author(s) 2018 Abstract Deep learning techniques have shown success in learning from raw high-dimensional data in various applications. While deep reinforcement learning is recently gaining popularity as a method to train intelligent agents, utilizing deep learning in imitation learning has been scarcely explored. Imitation learning can be an efficient method to teach intelligent agents by providing a set of demonstrations to learn from. However, generalizing to situations that are not represented in the demonstrations can be challenging, especially in 3D environments. In this paper, we propose a deep imitation learning method to learn navigation tasks from demonstrations in a 3D environment. The supervised policy is refined using active learning in order to generalize to unseen situations. This approach is compared to two popular deep reinforcement learning techniques: deep-Q-networks and Asynchronous actor-critic (A3C). The proposed method as well as the reinforcement learning methods employ deep convolutional neural networks and learn directly from raw visual input. Methods for combining learning from demonstrations and experience are also investigated. This combination aims to join the generalization ability of learning by experience with the efficiency of learning by imitation. The proposed methods are evaluated on 4 navigation tasks in a 3D simulated environment. Navigation tasks are a typical problem that is relevant to many real applications. They pose the challenge of requiring demonstrations of long trajectories to reach the target and only providing delayed rewards (usually terminal) to the agent. The experiments show that the proposed method can successfully learn navigation tasks from raw visual input while learning from experience methods fail to learn an effective policy. Moreover, it is shown that active learning can significantly improve the performance of the initially learned policy using a small number of active samples. Deep learning Convolutional neural networks Learning from demonstrations Reinforcement learning Active learning 3D navigation Benchmarking Elyan, Eyad aut Gaber, Mohamed Medhat aut Jayne, Chrisina aut Enthalten in Neural computing & applications Springer London, 1993 29(2017), 7 vom: 04. Dez., Seite 389-404 (DE-627)165669608 (DE-600)1136944-9 (DE-576)032873050 0941-0643 nnns volume:29 year:2017 number:7 day:04 month:12 pages:389-404 https://doi.org/10.1007/s00521-017-3241-z lizenzpflichtig Volltext GBV_USEFLAG_A SYSFLAG_A GBV_OLC SSG-OLC-MAT GBV_ILN_70 GBV_ILN_2018 GBV_ILN_4046 GBV_ILN_4277 AR 29 2017 7 04 12 389-404 |
language |
English |
source |
Enthalten in Neural computing & applications 29(2017), 7 vom: 04. Dez., Seite 389-404 volume:29 year:2017 number:7 day:04 month:12 pages:389-404 |
sourceStr |
Enthalten in Neural computing & applications 29(2017), 7 vom: 04. Dez., Seite 389-404 volume:29 year:2017 number:7 day:04 month:12 pages:389-404 |
format_phy_str_mv |
Article |
institution |
findex.gbv.de |
topic_facet |
Deep learning Convolutional neural networks Learning from demonstrations Reinforcement learning Active learning 3D navigation Benchmarking |
dewey-raw |
004 |
isfreeaccess_bool |
false |
container_title |
Neural computing & applications |
authorswithroles_txt_mv |
Hussein, Ahmed @@aut@@ Elyan, Eyad @@aut@@ Gaber, Mohamed Medhat @@aut@@ Jayne, Chrisina @@aut@@ |
publishDateDaySort_date |
2017-12-04T00:00:00Z |
hierarchy_top_id |
165669608 |
dewey-sort |
14 |
id |
OLC2025604963 |
language_de |
englisch |
fullrecord |
<?xml version="1.0" encoding="UTF-8"?><collection xmlns="http://www.loc.gov/MARC21/slim"><record><leader>01000caa a22002652 4500</leader><controlfield tag="001">OLC2025604963</controlfield><controlfield tag="003">DE-627</controlfield><controlfield tag="005">20230502114736.0</controlfield><controlfield tag="007">tu</controlfield><controlfield tag="008">200820s2017 xx ||||| 00| ||eng c</controlfield><datafield tag="024" ind1="7" ind2=" "><subfield code="a">10.1007/s00521-017-3241-z</subfield><subfield code="2">doi</subfield></datafield><datafield tag="035" ind1=" " ind2=" "><subfield code="a">(DE-627)OLC2025604963</subfield></datafield><datafield tag="035" ind1=" " ind2=" "><subfield code="a">(DE-He213)s00521-017-3241-z-p</subfield></datafield><datafield tag="040" ind1=" " ind2=" "><subfield code="a">DE-627</subfield><subfield code="b">ger</subfield><subfield code="c">DE-627</subfield><subfield code="e">rakwb</subfield></datafield><datafield tag="041" ind1=" " ind2=" "><subfield code="a">eng</subfield></datafield><datafield tag="082" ind1="0" ind2="4"><subfield code="a">004</subfield><subfield code="q">VZ</subfield></datafield><datafield tag="100" ind1="1" ind2=" "><subfield code="a">Hussein, Ahmed</subfield><subfield code="e">verfasserin</subfield><subfield code="0">(orcid)0000-0001-5227-9929</subfield><subfield code="4">aut</subfield></datafield><datafield tag="245" ind1="1" ind2="0"><subfield code="a">Deep imitation learning for 3D navigation tasks</subfield></datafield><datafield tag="264" ind1=" " ind2="1"><subfield code="c">2017</subfield></datafield><datafield tag="336" ind1=" " ind2=" "><subfield code="a">Text</subfield><subfield code="b">txt</subfield><subfield code="2">rdacontent</subfield></datafield><datafield tag="337" ind1=" " ind2=" "><subfield code="a">ohne Hilfsmittel zu benutzen</subfield><subfield code="b">n</subfield><subfield code="2">rdamedia</subfield></datafield><datafield tag="338" ind1=" " ind2=" "><subfield code="a">Band</subfield><subfield code="b">nc</subfield><subfield code="2">rdacarrier</subfield></datafield><datafield tag="500" ind1=" " ind2=" "><subfield code="a">© The Author(s) 2018</subfield></datafield><datafield tag="520" ind1=" " ind2=" "><subfield code="a">Abstract Deep learning techniques have shown success in learning from raw high-dimensional data in various applications. While deep reinforcement learning is recently gaining popularity as a method to train intelligent agents, utilizing deep learning in imitation learning has been scarcely explored. Imitation learning can be an efficient method to teach intelligent agents by providing a set of demonstrations to learn from. However, generalizing to situations that are not represented in the demonstrations can be challenging, especially in 3D environments. In this paper, we propose a deep imitation learning method to learn navigation tasks from demonstrations in a 3D environment. The supervised policy is refined using active learning in order to generalize to unseen situations. This approach is compared to two popular deep reinforcement learning techniques: deep-Q-networks and Asynchronous actor-critic (A3C). The proposed method as well as the reinforcement learning methods employ deep convolutional neural networks and learn directly from raw visual input. Methods for combining learning from demonstrations and experience are also investigated. This combination aims to join the generalization ability of learning by experience with the efficiency of learning by imitation. The proposed methods are evaluated on 4 navigation tasks in a 3D simulated environment. Navigation tasks are a typical problem that is relevant to many real applications. They pose the challenge of requiring demonstrations of long trajectories to reach the target and only providing delayed rewards (usually terminal) to the agent. The experiments show that the proposed method can successfully learn navigation tasks from raw visual input while learning from experience methods fail to learn an effective policy. Moreover, it is shown that active learning can significantly improve the performance of the initially learned policy using a small number of active samples.</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Deep learning</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Convolutional neural networks</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Learning from demonstrations</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Reinforcement learning</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Active learning</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">3D navigation</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Benchmarking</subfield></datafield><datafield tag="700" ind1="1" ind2=" "><subfield code="a">Elyan, Eyad</subfield><subfield code="4">aut</subfield></datafield><datafield tag="700" ind1="1" ind2=" "><subfield code="a">Gaber, Mohamed Medhat</subfield><subfield code="4">aut</subfield></datafield><datafield tag="700" ind1="1" ind2=" "><subfield code="a">Jayne, Chrisina</subfield><subfield code="4">aut</subfield></datafield><datafield tag="773" ind1="0" ind2="8"><subfield code="i">Enthalten in</subfield><subfield code="t">Neural computing & applications</subfield><subfield code="d">Springer London, 1993</subfield><subfield code="g">29(2017), 7 vom: 04. Dez., Seite 389-404</subfield><subfield code="w">(DE-627)165669608</subfield><subfield code="w">(DE-600)1136944-9</subfield><subfield code="w">(DE-576)032873050</subfield><subfield code="x">0941-0643</subfield><subfield code="7">nnns</subfield></datafield><datafield tag="773" ind1="1" ind2="8"><subfield code="g">volume:29</subfield><subfield code="g">year:2017</subfield><subfield code="g">number:7</subfield><subfield code="g">day:04</subfield><subfield code="g">month:12</subfield><subfield code="g">pages:389-404</subfield></datafield><datafield tag="856" ind1="4" ind2="1"><subfield code="u">https://doi.org/10.1007/s00521-017-3241-z</subfield><subfield code="z">lizenzpflichtig</subfield><subfield code="3">Volltext</subfield></datafield><datafield tag="912" ind1=" " ind2=" "><subfield code="a">GBV_USEFLAG_A</subfield></datafield><datafield tag="912" ind1=" " ind2=" "><subfield code="a">SYSFLAG_A</subfield></datafield><datafield tag="912" ind1=" " ind2=" "><subfield code="a">GBV_OLC</subfield></datafield><datafield tag="912" ind1=" " ind2=" "><subfield code="a">SSG-OLC-MAT</subfield></datafield><datafield tag="912" ind1=" " ind2=" "><subfield code="a">GBV_ILN_70</subfield></datafield><datafield tag="912" ind1=" " ind2=" "><subfield code="a">GBV_ILN_2018</subfield></datafield><datafield tag="912" ind1=" " ind2=" "><subfield code="a">GBV_ILN_4046</subfield></datafield><datafield tag="912" ind1=" " ind2=" "><subfield code="a">GBV_ILN_4277</subfield></datafield><datafield tag="951" ind1=" " ind2=" "><subfield code="a">AR</subfield></datafield><datafield tag="952" ind1=" " ind2=" "><subfield code="d">29</subfield><subfield code="j">2017</subfield><subfield code="e">7</subfield><subfield code="b">04</subfield><subfield code="c">12</subfield><subfield code="h">389-404</subfield></datafield></record></collection>
|
author |
Hussein, Ahmed |
spellingShingle |
Hussein, Ahmed ddc 004 misc Deep learning misc Convolutional neural networks misc Learning from demonstrations misc Reinforcement learning misc Active learning misc 3D navigation misc Benchmarking Deep imitation learning for 3D navigation tasks |
authorStr |
Hussein, Ahmed |
ppnlink_with_tag_str_mv |
@@773@@(DE-627)165669608 |
format |
Article |
dewey-ones |
004 - Data processing & computer science |
delete_txt_mv |
keep |
author_role |
aut aut aut aut |
collection |
OLC |
remote_str |
false |
illustrated |
Not Illustrated |
issn |
0941-0643 |
topic_title |
004 VZ Deep imitation learning for 3D navigation tasks Deep learning Convolutional neural networks Learning from demonstrations Reinforcement learning Active learning 3D navigation Benchmarking |
topic |
ddc 004 misc Deep learning misc Convolutional neural networks misc Learning from demonstrations misc Reinforcement learning misc Active learning misc 3D navigation misc Benchmarking |
topic_unstemmed |
ddc 004 misc Deep learning misc Convolutional neural networks misc Learning from demonstrations misc Reinforcement learning misc Active learning misc 3D navigation misc Benchmarking |
topic_browse |
ddc 004 misc Deep learning misc Convolutional neural networks misc Learning from demonstrations misc Reinforcement learning misc Active learning misc 3D navigation misc Benchmarking |
format_facet |
Aufsätze Gedruckte Aufsätze |
format_main_str_mv |
Text Zeitschrift/Artikel |
carriertype_str_mv |
nc |
hierarchy_parent_title |
Neural computing & applications |
hierarchy_parent_id |
165669608 |
dewey-tens |
000 - Computer science, knowledge & systems |
hierarchy_top_title |
Neural computing & applications |
isfreeaccess_txt |
false |
familylinks_str_mv |
(DE-627)165669608 (DE-600)1136944-9 (DE-576)032873050 |
title |
Deep imitation learning for 3D navigation tasks |
ctrlnum |
(DE-627)OLC2025604963 (DE-He213)s00521-017-3241-z-p |
title_full |
Deep imitation learning for 3D navigation tasks |
author_sort |
Hussein, Ahmed |
journal |
Neural computing & applications |
journalStr |
Neural computing & applications |
lang_code |
eng |
isOA_bool |
false |
dewey-hundreds |
000 - Computer science, information & general works |
recordtype |
marc |
publishDateSort |
2017 |
contenttype_str_mv |
txt |
container_start_page |
389 |
author_browse |
Hussein, Ahmed Elyan, Eyad Gaber, Mohamed Medhat Jayne, Chrisina |
container_volume |
29 |
class |
004 VZ |
format_se |
Aufsätze |
author-letter |
Hussein, Ahmed |
doi_str_mv |
10.1007/s00521-017-3241-z |
normlink |
(ORCID)0000-0001-5227-9929 |
normlink_prefix_str_mv |
(orcid)0000-0001-5227-9929 |
dewey-full |
004 |
title_sort |
deep imitation learning for 3d navigation tasks |
title_auth |
Deep imitation learning for 3D navigation tasks |
abstract |
Abstract Deep learning techniques have shown success in learning from raw high-dimensional data in various applications. While deep reinforcement learning is recently gaining popularity as a method to train intelligent agents, utilizing deep learning in imitation learning has been scarcely explored. Imitation learning can be an efficient method to teach intelligent agents by providing a set of demonstrations to learn from. However, generalizing to situations that are not represented in the demonstrations can be challenging, especially in 3D environments. In this paper, we propose a deep imitation learning method to learn navigation tasks from demonstrations in a 3D environment. The supervised policy is refined using active learning in order to generalize to unseen situations. This approach is compared to two popular deep reinforcement learning techniques: deep-Q-networks and Asynchronous actor-critic (A3C). The proposed method as well as the reinforcement learning methods employ deep convolutional neural networks and learn directly from raw visual input. Methods for combining learning from demonstrations and experience are also investigated. This combination aims to join the generalization ability of learning by experience with the efficiency of learning by imitation. The proposed methods are evaluated on 4 navigation tasks in a 3D simulated environment. Navigation tasks are a typical problem that is relevant to many real applications. They pose the challenge of requiring demonstrations of long trajectories to reach the target and only providing delayed rewards (usually terminal) to the agent. The experiments show that the proposed method can successfully learn navigation tasks from raw visual input while learning from experience methods fail to learn an effective policy. Moreover, it is shown that active learning can significantly improve the performance of the initially learned policy using a small number of active samples. © The Author(s) 2018 |
abstractGer |
Abstract Deep learning techniques have shown success in learning from raw high-dimensional data in various applications. While deep reinforcement learning is recently gaining popularity as a method to train intelligent agents, utilizing deep learning in imitation learning has been scarcely explored. Imitation learning can be an efficient method to teach intelligent agents by providing a set of demonstrations to learn from. However, generalizing to situations that are not represented in the demonstrations can be challenging, especially in 3D environments. In this paper, we propose a deep imitation learning method to learn navigation tasks from demonstrations in a 3D environment. The supervised policy is refined using active learning in order to generalize to unseen situations. This approach is compared to two popular deep reinforcement learning techniques: deep-Q-networks and Asynchronous actor-critic (A3C). The proposed method as well as the reinforcement learning methods employ deep convolutional neural networks and learn directly from raw visual input. Methods for combining learning from demonstrations and experience are also investigated. This combination aims to join the generalization ability of learning by experience with the efficiency of learning by imitation. The proposed methods are evaluated on 4 navigation tasks in a 3D simulated environment. Navigation tasks are a typical problem that is relevant to many real applications. They pose the challenge of requiring demonstrations of long trajectories to reach the target and only providing delayed rewards (usually terminal) to the agent. The experiments show that the proposed method can successfully learn navigation tasks from raw visual input while learning from experience methods fail to learn an effective policy. Moreover, it is shown that active learning can significantly improve the performance of the initially learned policy using a small number of active samples. © The Author(s) 2018 |
abstract_unstemmed |
Abstract Deep learning techniques have shown success in learning from raw high-dimensional data in various applications. While deep reinforcement learning is recently gaining popularity as a method to train intelligent agents, utilizing deep learning in imitation learning has been scarcely explored. Imitation learning can be an efficient method to teach intelligent agents by providing a set of demonstrations to learn from. However, generalizing to situations that are not represented in the demonstrations can be challenging, especially in 3D environments. In this paper, we propose a deep imitation learning method to learn navigation tasks from demonstrations in a 3D environment. The supervised policy is refined using active learning in order to generalize to unseen situations. This approach is compared to two popular deep reinforcement learning techniques: deep-Q-networks and Asynchronous actor-critic (A3C). The proposed method as well as the reinforcement learning methods employ deep convolutional neural networks and learn directly from raw visual input. Methods for combining learning from demonstrations and experience are also investigated. This combination aims to join the generalization ability of learning by experience with the efficiency of learning by imitation. The proposed methods are evaluated on 4 navigation tasks in a 3D simulated environment. Navigation tasks are a typical problem that is relevant to many real applications. They pose the challenge of requiring demonstrations of long trajectories to reach the target and only providing delayed rewards (usually terminal) to the agent. The experiments show that the proposed method can successfully learn navigation tasks from raw visual input while learning from experience methods fail to learn an effective policy. Moreover, it is shown that active learning can significantly improve the performance of the initially learned policy using a small number of active samples. © The Author(s) 2018 |
collection_details |
GBV_USEFLAG_A SYSFLAG_A GBV_OLC SSG-OLC-MAT GBV_ILN_70 GBV_ILN_2018 GBV_ILN_4046 GBV_ILN_4277 |
container_issue |
7 |
title_short |
Deep imitation learning for 3D navigation tasks |
url |
https://doi.org/10.1007/s00521-017-3241-z |
remote_bool |
false |
author2 |
Elyan, Eyad Gaber, Mohamed Medhat Jayne, Chrisina |
author2Str |
Elyan, Eyad Gaber, Mohamed Medhat Jayne, Chrisina |
ppnlink |
165669608 |
mediatype_str_mv |
n |
isOA_txt |
false |
hochschulschrift_bool |
false |
doi_str |
10.1007/s00521-017-3241-z |
up_date |
2024-07-04T01:40:53.908Z |
_version_ |
1803610755073835008 |
fullrecord_marcxml |
<?xml version="1.0" encoding="UTF-8"?><collection xmlns="http://www.loc.gov/MARC21/slim"><record><leader>01000caa a22002652 4500</leader><controlfield tag="001">OLC2025604963</controlfield><controlfield tag="003">DE-627</controlfield><controlfield tag="005">20230502114736.0</controlfield><controlfield tag="007">tu</controlfield><controlfield tag="008">200820s2017 xx ||||| 00| ||eng c</controlfield><datafield tag="024" ind1="7" ind2=" "><subfield code="a">10.1007/s00521-017-3241-z</subfield><subfield code="2">doi</subfield></datafield><datafield tag="035" ind1=" " ind2=" "><subfield code="a">(DE-627)OLC2025604963</subfield></datafield><datafield tag="035" ind1=" " ind2=" "><subfield code="a">(DE-He213)s00521-017-3241-z-p</subfield></datafield><datafield tag="040" ind1=" " ind2=" "><subfield code="a">DE-627</subfield><subfield code="b">ger</subfield><subfield code="c">DE-627</subfield><subfield code="e">rakwb</subfield></datafield><datafield tag="041" ind1=" " ind2=" "><subfield code="a">eng</subfield></datafield><datafield tag="082" ind1="0" ind2="4"><subfield code="a">004</subfield><subfield code="q">VZ</subfield></datafield><datafield tag="100" ind1="1" ind2=" "><subfield code="a">Hussein, Ahmed</subfield><subfield code="e">verfasserin</subfield><subfield code="0">(orcid)0000-0001-5227-9929</subfield><subfield code="4">aut</subfield></datafield><datafield tag="245" ind1="1" ind2="0"><subfield code="a">Deep imitation learning for 3D navigation tasks</subfield></datafield><datafield tag="264" ind1=" " ind2="1"><subfield code="c">2017</subfield></datafield><datafield tag="336" ind1=" " ind2=" "><subfield code="a">Text</subfield><subfield code="b">txt</subfield><subfield code="2">rdacontent</subfield></datafield><datafield tag="337" ind1=" " ind2=" "><subfield code="a">ohne Hilfsmittel zu benutzen</subfield><subfield code="b">n</subfield><subfield code="2">rdamedia</subfield></datafield><datafield tag="338" ind1=" " ind2=" "><subfield code="a">Band</subfield><subfield code="b">nc</subfield><subfield code="2">rdacarrier</subfield></datafield><datafield tag="500" ind1=" " ind2=" "><subfield code="a">© The Author(s) 2018</subfield></datafield><datafield tag="520" ind1=" " ind2=" "><subfield code="a">Abstract Deep learning techniques have shown success in learning from raw high-dimensional data in various applications. While deep reinforcement learning is recently gaining popularity as a method to train intelligent agents, utilizing deep learning in imitation learning has been scarcely explored. Imitation learning can be an efficient method to teach intelligent agents by providing a set of demonstrations to learn from. However, generalizing to situations that are not represented in the demonstrations can be challenging, especially in 3D environments. In this paper, we propose a deep imitation learning method to learn navigation tasks from demonstrations in a 3D environment. The supervised policy is refined using active learning in order to generalize to unseen situations. This approach is compared to two popular deep reinforcement learning techniques: deep-Q-networks and Asynchronous actor-critic (A3C). The proposed method as well as the reinforcement learning methods employ deep convolutional neural networks and learn directly from raw visual input. Methods for combining learning from demonstrations and experience are also investigated. This combination aims to join the generalization ability of learning by experience with the efficiency of learning by imitation. The proposed methods are evaluated on 4 navigation tasks in a 3D simulated environment. Navigation tasks are a typical problem that is relevant to many real applications. They pose the challenge of requiring demonstrations of long trajectories to reach the target and only providing delayed rewards (usually terminal) to the agent. The experiments show that the proposed method can successfully learn navigation tasks from raw visual input while learning from experience methods fail to learn an effective policy. Moreover, it is shown that active learning can significantly improve the performance of the initially learned policy using a small number of active samples.</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Deep learning</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Convolutional neural networks</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Learning from demonstrations</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Reinforcement learning</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Active learning</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">3D navigation</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Benchmarking</subfield></datafield><datafield tag="700" ind1="1" ind2=" "><subfield code="a">Elyan, Eyad</subfield><subfield code="4">aut</subfield></datafield><datafield tag="700" ind1="1" ind2=" "><subfield code="a">Gaber, Mohamed Medhat</subfield><subfield code="4">aut</subfield></datafield><datafield tag="700" ind1="1" ind2=" "><subfield code="a">Jayne, Chrisina</subfield><subfield code="4">aut</subfield></datafield><datafield tag="773" ind1="0" ind2="8"><subfield code="i">Enthalten in</subfield><subfield code="t">Neural computing & applications</subfield><subfield code="d">Springer London, 1993</subfield><subfield code="g">29(2017), 7 vom: 04. Dez., Seite 389-404</subfield><subfield code="w">(DE-627)165669608</subfield><subfield code="w">(DE-600)1136944-9</subfield><subfield code="w">(DE-576)032873050</subfield><subfield code="x">0941-0643</subfield><subfield code="7">nnns</subfield></datafield><datafield tag="773" ind1="1" ind2="8"><subfield code="g">volume:29</subfield><subfield code="g">year:2017</subfield><subfield code="g">number:7</subfield><subfield code="g">day:04</subfield><subfield code="g">month:12</subfield><subfield code="g">pages:389-404</subfield></datafield><datafield tag="856" ind1="4" ind2="1"><subfield code="u">https://doi.org/10.1007/s00521-017-3241-z</subfield><subfield code="z">lizenzpflichtig</subfield><subfield code="3">Volltext</subfield></datafield><datafield tag="912" ind1=" " ind2=" "><subfield code="a">GBV_USEFLAG_A</subfield></datafield><datafield tag="912" ind1=" " ind2=" "><subfield code="a">SYSFLAG_A</subfield></datafield><datafield tag="912" ind1=" " ind2=" "><subfield code="a">GBV_OLC</subfield></datafield><datafield tag="912" ind1=" " ind2=" "><subfield code="a">SSG-OLC-MAT</subfield></datafield><datafield tag="912" ind1=" " ind2=" "><subfield code="a">GBV_ILN_70</subfield></datafield><datafield tag="912" ind1=" " ind2=" "><subfield code="a">GBV_ILN_2018</subfield></datafield><datafield tag="912" ind1=" " ind2=" "><subfield code="a">GBV_ILN_4046</subfield></datafield><datafield tag="912" ind1=" " ind2=" "><subfield code="a">GBV_ILN_4277</subfield></datafield><datafield tag="951" ind1=" " ind2=" "><subfield code="a">AR</subfield></datafield><datafield tag="952" ind1=" " ind2=" "><subfield code="d">29</subfield><subfield code="j">2017</subfield><subfield code="e">7</subfield><subfield code="b">04</subfield><subfield code="c">12</subfield><subfield code="h">389-404</subfield></datafield></record></collection>
|
score |
7.4009476 |