Random Forest Construction With Robust Semisupervised Node Splitting

Random forest (RF) is a very important classifier with applications in various machine learning tasks, but its promising performance heavily relies on the size of labeled training data. In this paper, we investigate constructing of RFs with a small size of labeled data and find that the performance...
Ausführliche Beschreibung

Gespeichert in:

Autor*in:	Xiao Liu [verfasserIn] Mingli Song Dacheng Tao Zicheng Liu Luming Zhang Chun Chen Jiajun Bu

Format:	Artikel
Sprache:	Englisch

Erschienen:	2015

Schlagwörter:	low-dimensional subspace learning (artificial intelligence) Covariance matrices semi-supervised splitting local maxima Estimation semi-supervised learning random forest random forest construction hyperplane separation kernel-based density estimation robust semisupervised node splitting labeled training data Training data unified optimization framework Kernel data points asymptotic mean integrated squared error criterion image classification Bandwidth computer vision Training supervised algorithms high-dimensional feature space Radio frequency subspace learning machine learning tasks quality measure margin maximization-based semisupervised methods node splitting optimisation Feature extraction Face - anatomy & histology Image Processing, Computer-Assisted - methods Pattern Recognition, Automated - methods

Übergeordnetes Werk:	Enthalten in: IEEE transactions on image processing - New York, NY : Inst., 1992, 24(2015), 1, Seite 471-483
Übergeordnetes Werk:	volume:24 ; year:2015 ; number:1 ; pages:471-483

Links:	Volltext Link aufrufen Link aufrufen Link aufrufen

DOI / URN:	10.1109/TIP.2014.2378017

Katalog-ID:	OLC1959239910

Internformat


LEADER	01000caa a2200265 4500
001	OLC1959239910
003	DE-627
005	20230714150846.0
007	tu
008	160206s2015 xx \|\|\|\|\| 00\| \|\|eng c
024	7		\|a 10.1109/TIP.2014.2378017 \|2 doi
028	5	2	\|a PQ20160617
035			\|a (DE-627)OLC1959239910
035			\|a (DE-599)GBVOLC1959239910
035			\|a (PRQ)c2625-33afd5198206abf0a23259fccd5db890d7a414e7e5caf12f44ca42b8d094e2120
035			\|a (KEY)0213811520150000024000100471randomforestconstructionwithrobustsemisupervisedno
040			\|a DE-627 \|b ger \|c DE-627 \|e rakwb
041			\|a eng
082	0	4	\|a 004 \|a 620 \|q DNB
084			\|a 54.00 \|2 bkl
100	0		\|a Xiao Liu \|e verfasserin \|4 aut
245	1	0	\|a Random Forest Construction With Robust Semisupervised Node Splitting
264		1	\|c 2015
336			\|a Text \|b txt \|2 rdacontent
337			\|a ohne Hilfsmittel zu benutzen \|b n \|2 rdamedia
338			\|a Band \|b nc \|2 rdacarrier
520			\|a Random forest (RF) is a very important classifier with applications in various machine learning tasks, but its promising performance heavily relies on the size of labeled training data. In this paper, we investigate constructing of RFs with a small size of labeled data and find that the performance bottleneck is located in the node splitting procedures; hence, existing solutions fail to properly partition the feature space if there are insufficient training data. To achieve robust node splitting with insufficient data, we present semisupervised splitting to overcome this limitation by splitting nodes with the guidance of both labeled and abundant unlabeled data. In particular, an accurate quality measure of node splitting is obtained by carrying out the kernel-based density estimation, whereby a multiclass version of asymptotic mean integrated squared error criterion is proposed to adaptively select the optimal bandwidth of the kernel. To avoid the curse of dimensionality, we project the data points from the original high-dimensional feature space onto a low-dimensional subspace before estimation. A unified optimization framework is proposed to select a coupled pair of subspace and separating hyperplane such that the smoothness of the subspace and the quality of the splitting are guaranteed simultaneously. Our algorithm efficiently avoids overfitting caused by bad initialization and local maxima when compared with conventional margin maximization-based semisupervised methods. We demonstrate the effectiveness of the proposed algorithm by comparing it with state-of-the-art supervised and semisupervised algorithms for typical computer vision applications, such as object categorization, face recognition, and image segmentation, on publicly available data sets.
650		4	\|a low-dimensional subspace
650		4	\|a learning (artificial intelligence)
650		4	\|a Covariance matrices
650		4	\|a semi-supervised splitting
650		4	\|a local maxima
650		4	\|a Estimation
650		4	\|a semi-supervised learning
650		4	\|a random forest
650		4	\|a random forest construction
650		4	\|a hyperplane separation
650		4	\|a kernel-based density estimation
650		4	\|a robust semisupervised node splitting
650		4	\|a labeled training data
650		4	\|a Training data
650		4	\|a unified optimization framework
650		4	\|a Kernel
650		4	\|a data points
650		4	\|a asymptotic mean integrated squared error criterion
650		4	\|a image classification
650		4	\|a Bandwidth
650		4	\|a computer vision
650		4	\|a Training
650		4	\|a supervised algorithms
650		4	\|a high-dimensional feature space
650		4	\|a Radio frequency
650		4	\|a subspace learning
650		4	\|a machine learning tasks
650		4	\|a quality measure
650		4	\|a margin maximization-based semisupervised methods
650		4	\|a node splitting
650		4	\|a optimisation
650		4	\|a Feature extraction
650		4	\|a Face - anatomy & histology
650		4	\|a Image Processing, Computer-Assisted - methods
650		4	\|a Pattern Recognition, Automated - methods
700	0		\|a Mingli Song \|4 oth
700	0		\|a Dacheng Tao \|4 oth
700	0		\|a Zicheng Liu \|4 oth
700	0		\|a Luming Zhang \|4 oth
700	0		\|a Chun Chen \|4 oth
700	0		\|a Jiajun Bu \|4 oth
773	0	8	\|i Enthalten in \|t IEEE transactions on image processing \|d New York, NY : Inst., 1992 \|g 24(2015), 1, Seite 471-483 \|w (DE-627)131074458 \|w (DE-600)1111265-7 \|w (DE-576)029165008 \|x 1057-7149 \|7 nnns
773	1	8	\|g volume:24 \|g year:2015 \|g number:1 \|g pages:471-483
856	4	1	\|u http://dx.doi.org/10.1109/TIP.2014.2378017 \|3 Volltext
856	4	2	\|u http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=6975199
856	4	2	\|u http://www.ncbi.nlm.nih.gov/pubmed/25494503
856	4	2	\|u http://search.proquest.com/docview/1640798292
912			\|a GBV_USEFLAG_A
912			\|a SYSFLAG_A
912			\|a GBV_OLC
912			\|a SSG-OLC-TEC
912			\|a SSG-OLC-MAT
912			\|a GBV_ILN_70
912			\|a GBV_ILN_2005
912			\|a GBV_ILN_4318
936	b	k	\|a 54.00 \|q AVZ
951			\|a AR
952			\|d 24 \|j 2015 \|e 1 \|h 471-483

Indexfelder

author_variant	x l xl
matchkey_str	article:10577149:2015----::admoetosrcinihouteiuev
hierarchy_sort_str	2015
bklnumber	54.00
publishDate	2015
allfields	10.1109/TIP.2014.2378017 doi PQ20160617 (DE-627)OLC1959239910 (DE-599)GBVOLC1959239910 (PRQ)c2625-33afd5198206abf0a23259fccd5db890d7a414e7e5caf12f44ca42b8d094e2120 (KEY)0213811520150000024000100471randomforestconstructionwithrobustsemisupervisedno DE-627 ger DE-627 rakwb eng 004 620 DNB 54.00 bkl Xiao Liu verfasserin aut Random Forest Construction With Robust Semisupervised Node Splitting 2015 Text txt rdacontent ohne Hilfsmittel zu benutzen n rdamedia Band nc rdacarrier Random forest (RF) is a very important classifier with applications in various machine learning tasks, but its promising performance heavily relies on the size of labeled training data. In this paper, we investigate constructing of RFs with a small size of labeled data and find that the performance bottleneck is located in the node splitting procedures; hence, existing solutions fail to properly partition the feature space if there are insufficient training data. To achieve robust node splitting with insufficient data, we present semisupervised splitting to overcome this limitation by splitting nodes with the guidance of both labeled and abundant unlabeled data. In particular, an accurate quality measure of node splitting is obtained by carrying out the kernel-based density estimation, whereby a multiclass version of asymptotic mean integrated squared error criterion is proposed to adaptively select the optimal bandwidth of the kernel. To avoid the curse of dimensionality, we project the data points from the original high-dimensional feature space onto a low-dimensional subspace before estimation. A unified optimization framework is proposed to select a coupled pair of subspace and separating hyperplane such that the smoothness of the subspace and the quality of the splitting are guaranteed simultaneously. Our algorithm efficiently avoids overfitting caused by bad initialization and local maxima when compared with conventional margin maximization-based semisupervised methods. We demonstrate the effectiveness of the proposed algorithm by comparing it with state-of-the-art supervised and semisupervised algorithms for typical computer vision applications, such as object categorization, face recognition, and image segmentation, on publicly available data sets. low-dimensional subspace learning (artificial intelligence) Covariance matrices semi-supervised splitting local maxima Estimation semi-supervised learning random forest random forest construction hyperplane separation kernel-based density estimation robust semisupervised node splitting labeled training data Training data unified optimization framework Kernel data points asymptotic mean integrated squared error criterion image classification Bandwidth computer vision Training supervised algorithms high-dimensional feature space Radio frequency subspace learning machine learning tasks quality measure margin maximization-based semisupervised methods node splitting optimisation Feature extraction Face - anatomy & histology Image Processing, Computer-Assisted - methods Pattern Recognition, Automated - methods Mingli Song oth Dacheng Tao oth Zicheng Liu oth Luming Zhang oth Chun Chen oth Jiajun Bu oth Enthalten in IEEE transactions on image processing New York, NY : Inst., 1992 24(2015), 1, Seite 471-483 (DE-627)131074458 (DE-600)1111265-7 (DE-576)029165008 1057-7149 nnns volume:24 year:2015 number:1 pages:471-483 http://dx.doi.org/10.1109/TIP.2014.2378017 Volltext http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=6975199 http://www.ncbi.nlm.nih.gov/pubmed/25494503 http://search.proquest.com/docview/1640798292 GBV_USEFLAG_A SYSFLAG_A GBV_OLC SSG-OLC-TEC SSG-OLC-MAT GBV_ILN_70 GBV_ILN_2005 GBV_ILN_4318 54.00 AVZ AR 24 2015 1 471-483
spelling	10.1109/TIP.2014.2378017 doi PQ20160617 (DE-627)OLC1959239910 (DE-599)GBVOLC1959239910 (PRQ)c2625-33afd5198206abf0a23259fccd5db890d7a414e7e5caf12f44ca42b8d094e2120 (KEY)0213811520150000024000100471randomforestconstructionwithrobustsemisupervisedno DE-627 ger DE-627 rakwb eng 004 620 DNB 54.00 bkl Xiao Liu verfasserin aut Random Forest Construction With Robust Semisupervised Node Splitting 2015 Text txt rdacontent ohne Hilfsmittel zu benutzen n rdamedia Band nc rdacarrier Random forest (RF) is a very important classifier with applications in various machine learning tasks, but its promising performance heavily relies on the size of labeled training data. In this paper, we investigate constructing of RFs with a small size of labeled data and find that the performance bottleneck is located in the node splitting procedures; hence, existing solutions fail to properly partition the feature space if there are insufficient training data. To achieve robust node splitting with insufficient data, we present semisupervised splitting to overcome this limitation by splitting nodes with the guidance of both labeled and abundant unlabeled data. In particular, an accurate quality measure of node splitting is obtained by carrying out the kernel-based density estimation, whereby a multiclass version of asymptotic mean integrated squared error criterion is proposed to adaptively select the optimal bandwidth of the kernel. To avoid the curse of dimensionality, we project the data points from the original high-dimensional feature space onto a low-dimensional subspace before estimation. A unified optimization framework is proposed to select a coupled pair of subspace and separating hyperplane such that the smoothness of the subspace and the quality of the splitting are guaranteed simultaneously. Our algorithm efficiently avoids overfitting caused by bad initialization and local maxima when compared with conventional margin maximization-based semisupervised methods. We demonstrate the effectiveness of the proposed algorithm by comparing it with state-of-the-art supervised and semisupervised algorithms for typical computer vision applications, such as object categorization, face recognition, and image segmentation, on publicly available data sets. low-dimensional subspace learning (artificial intelligence) Covariance matrices semi-supervised splitting local maxima Estimation semi-supervised learning random forest random forest construction hyperplane separation kernel-based density estimation robust semisupervised node splitting labeled training data Training data unified optimization framework Kernel data points asymptotic mean integrated squared error criterion image classification Bandwidth computer vision Training supervised algorithms high-dimensional feature space Radio frequency subspace learning machine learning tasks quality measure margin maximization-based semisupervised methods node splitting optimisation Feature extraction Face - anatomy & histology Image Processing, Computer-Assisted - methods Pattern Recognition, Automated - methods Mingli Song oth Dacheng Tao oth Zicheng Liu oth Luming Zhang oth Chun Chen oth Jiajun Bu oth Enthalten in IEEE transactions on image processing New York, NY : Inst., 1992 24(2015), 1, Seite 471-483 (DE-627)131074458 (DE-600)1111265-7 (DE-576)029165008 1057-7149 nnns volume:24 year:2015 number:1 pages:471-483 http://dx.doi.org/10.1109/TIP.2014.2378017 Volltext http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=6975199 http://www.ncbi.nlm.nih.gov/pubmed/25494503 http://search.proquest.com/docview/1640798292 GBV_USEFLAG_A SYSFLAG_A GBV_OLC SSG-OLC-TEC SSG-OLC-MAT GBV_ILN_70 GBV_ILN_2005 GBV_ILN_4318 54.00 AVZ AR 24 2015 1 471-483
allfields_unstemmed	10.1109/TIP.2014.2378017 doi PQ20160617 (DE-627)OLC1959239910 (DE-599)GBVOLC1959239910 (PRQ)c2625-33afd5198206abf0a23259fccd5db890d7a414e7e5caf12f44ca42b8d094e2120 (KEY)0213811520150000024000100471randomforestconstructionwithrobustsemisupervisedno DE-627 ger DE-627 rakwb eng 004 620 DNB 54.00 bkl Xiao Liu verfasserin aut Random Forest Construction With Robust Semisupervised Node Splitting 2015 Text txt rdacontent ohne Hilfsmittel zu benutzen n rdamedia Band nc rdacarrier Random forest (RF) is a very important classifier with applications in various machine learning tasks, but its promising performance heavily relies on the size of labeled training data. In this paper, we investigate constructing of RFs with a small size of labeled data and find that the performance bottleneck is located in the node splitting procedures; hence, existing solutions fail to properly partition the feature space if there are insufficient training data. To achieve robust node splitting with insufficient data, we present semisupervised splitting to overcome this limitation by splitting nodes with the guidance of both labeled and abundant unlabeled data. In particular, an accurate quality measure of node splitting is obtained by carrying out the kernel-based density estimation, whereby a multiclass version of asymptotic mean integrated squared error criterion is proposed to adaptively select the optimal bandwidth of the kernel. To avoid the curse of dimensionality, we project the data points from the original high-dimensional feature space onto a low-dimensional subspace before estimation. A unified optimization framework is proposed to select a coupled pair of subspace and separating hyperplane such that the smoothness of the subspace and the quality of the splitting are guaranteed simultaneously. Our algorithm efficiently avoids overfitting caused by bad initialization and local maxima when compared with conventional margin maximization-based semisupervised methods. We demonstrate the effectiveness of the proposed algorithm by comparing it with state-of-the-art supervised and semisupervised algorithms for typical computer vision applications, such as object categorization, face recognition, and image segmentation, on publicly available data sets. low-dimensional subspace learning (artificial intelligence) Covariance matrices semi-supervised splitting local maxima Estimation semi-supervised learning random forest random forest construction hyperplane separation kernel-based density estimation robust semisupervised node splitting labeled training data Training data unified optimization framework Kernel data points asymptotic mean integrated squared error criterion image classification Bandwidth computer vision Training supervised algorithms high-dimensional feature space Radio frequency subspace learning machine learning tasks quality measure margin maximization-based semisupervised methods node splitting optimisation Feature extraction Face - anatomy & histology Image Processing, Computer-Assisted - methods Pattern Recognition, Automated - methods Mingli Song oth Dacheng Tao oth Zicheng Liu oth Luming Zhang oth Chun Chen oth Jiajun Bu oth Enthalten in IEEE transactions on image processing New York, NY : Inst., 1992 24(2015), 1, Seite 471-483 (DE-627)131074458 (DE-600)1111265-7 (DE-576)029165008 1057-7149 nnns volume:24 year:2015 number:1 pages:471-483 http://dx.doi.org/10.1109/TIP.2014.2378017 Volltext http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=6975199 http://www.ncbi.nlm.nih.gov/pubmed/25494503 http://search.proquest.com/docview/1640798292 GBV_USEFLAG_A SYSFLAG_A GBV_OLC SSG-OLC-TEC SSG-OLC-MAT GBV_ILN_70 GBV_ILN_2005 GBV_ILN_4318 54.00 AVZ AR 24 2015 1 471-483
allfieldsGer	10.1109/TIP.2014.2378017 doi PQ20160617 (DE-627)OLC1959239910 (DE-599)GBVOLC1959239910 (PRQ)c2625-33afd5198206abf0a23259fccd5db890d7a414e7e5caf12f44ca42b8d094e2120 (KEY)0213811520150000024000100471randomforestconstructionwithrobustsemisupervisedno DE-627 ger DE-627 rakwb eng 004 620 DNB 54.00 bkl Xiao Liu verfasserin aut Random Forest Construction With Robust Semisupervised Node Splitting 2015 Text txt rdacontent ohne Hilfsmittel zu benutzen n rdamedia Band nc rdacarrier Random forest (RF) is a very important classifier with applications in various machine learning tasks, but its promising performance heavily relies on the size of labeled training data. In this paper, we investigate constructing of RFs with a small size of labeled data and find that the performance bottleneck is located in the node splitting procedures; hence, existing solutions fail to properly partition the feature space if there are insufficient training data. To achieve robust node splitting with insufficient data, we present semisupervised splitting to overcome this limitation by splitting nodes with the guidance of both labeled and abundant unlabeled data. In particular, an accurate quality measure of node splitting is obtained by carrying out the kernel-based density estimation, whereby a multiclass version of asymptotic mean integrated squared error criterion is proposed to adaptively select the optimal bandwidth of the kernel. To avoid the curse of dimensionality, we project the data points from the original high-dimensional feature space onto a low-dimensional subspace before estimation. A unified optimization framework is proposed to select a coupled pair of subspace and separating hyperplane such that the smoothness of the subspace and the quality of the splitting are guaranteed simultaneously. Our algorithm efficiently avoids overfitting caused by bad initialization and local maxima when compared with conventional margin maximization-based semisupervised methods. We demonstrate the effectiveness of the proposed algorithm by comparing it with state-of-the-art supervised and semisupervised algorithms for typical computer vision applications, such as object categorization, face recognition, and image segmentation, on publicly available data sets. low-dimensional subspace learning (artificial intelligence) Covariance matrices semi-supervised splitting local maxima Estimation semi-supervised learning random forest random forest construction hyperplane separation kernel-based density estimation robust semisupervised node splitting labeled training data Training data unified optimization framework Kernel data points asymptotic mean integrated squared error criterion image classification Bandwidth computer vision Training supervised algorithms high-dimensional feature space Radio frequency subspace learning machine learning tasks quality measure margin maximization-based semisupervised methods node splitting optimisation Feature extraction Face - anatomy & histology Image Processing, Computer-Assisted - methods Pattern Recognition, Automated - methods Mingli Song oth Dacheng Tao oth Zicheng Liu oth Luming Zhang oth Chun Chen oth Jiajun Bu oth Enthalten in IEEE transactions on image processing New York, NY : Inst., 1992 24(2015), 1, Seite 471-483 (DE-627)131074458 (DE-600)1111265-7 (DE-576)029165008 1057-7149 nnns volume:24 year:2015 number:1 pages:471-483 http://dx.doi.org/10.1109/TIP.2014.2378017 Volltext http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=6975199 http://www.ncbi.nlm.nih.gov/pubmed/25494503 http://search.proquest.com/docview/1640798292 GBV_USEFLAG_A SYSFLAG_A GBV_OLC SSG-OLC-TEC SSG-OLC-MAT GBV_ILN_70 GBV_ILN_2005 GBV_ILN_4318 54.00 AVZ AR 24 2015 1 471-483
allfieldsSound	10.1109/TIP.2014.2378017 doi PQ20160617 (DE-627)OLC1959239910 (DE-599)GBVOLC1959239910 (PRQ)c2625-33afd5198206abf0a23259fccd5db890d7a414e7e5caf12f44ca42b8d094e2120 (KEY)0213811520150000024000100471randomforestconstructionwithrobustsemisupervisedno DE-627 ger DE-627 rakwb eng 004 620 DNB 54.00 bkl Xiao Liu verfasserin aut Random Forest Construction With Robust Semisupervised Node Splitting 2015 Text txt rdacontent ohne Hilfsmittel zu benutzen n rdamedia Band nc rdacarrier Random forest (RF) is a very important classifier with applications in various machine learning tasks, but its promising performance heavily relies on the size of labeled training data. In this paper, we investigate constructing of RFs with a small size of labeled data and find that the performance bottleneck is located in the node splitting procedures; hence, existing solutions fail to properly partition the feature space if there are insufficient training data. To achieve robust node splitting with insufficient data, we present semisupervised splitting to overcome this limitation by splitting nodes with the guidance of both labeled and abundant unlabeled data. In particular, an accurate quality measure of node splitting is obtained by carrying out the kernel-based density estimation, whereby a multiclass version of asymptotic mean integrated squared error criterion is proposed to adaptively select the optimal bandwidth of the kernel. To avoid the curse of dimensionality, we project the data points from the original high-dimensional feature space onto a low-dimensional subspace before estimation. A unified optimization framework is proposed to select a coupled pair of subspace and separating hyperplane such that the smoothness of the subspace and the quality of the splitting are guaranteed simultaneously. Our algorithm efficiently avoids overfitting caused by bad initialization and local maxima when compared with conventional margin maximization-based semisupervised methods. We demonstrate the effectiveness of the proposed algorithm by comparing it with state-of-the-art supervised and semisupervised algorithms for typical computer vision applications, such as object categorization, face recognition, and image segmentation, on publicly available data sets. low-dimensional subspace learning (artificial intelligence) Covariance matrices semi-supervised splitting local maxima Estimation semi-supervised learning random forest random forest construction hyperplane separation kernel-based density estimation robust semisupervised node splitting labeled training data Training data unified optimization framework Kernel data points asymptotic mean integrated squared error criterion image classification Bandwidth computer vision Training supervised algorithms high-dimensional feature space Radio frequency subspace learning machine learning tasks quality measure margin maximization-based semisupervised methods node splitting optimisation Feature extraction Face - anatomy & histology Image Processing, Computer-Assisted - methods Pattern Recognition, Automated - methods Mingli Song oth Dacheng Tao oth Zicheng Liu oth Luming Zhang oth Chun Chen oth Jiajun Bu oth Enthalten in IEEE transactions on image processing New York, NY : Inst., 1992 24(2015), 1, Seite 471-483 (DE-627)131074458 (DE-600)1111265-7 (DE-576)029165008 1057-7149 nnns volume:24 year:2015 number:1 pages:471-483 http://dx.doi.org/10.1109/TIP.2014.2378017 Volltext http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=6975199 http://www.ncbi.nlm.nih.gov/pubmed/25494503 http://search.proquest.com/docview/1640798292 GBV_USEFLAG_A SYSFLAG_A GBV_OLC SSG-OLC-TEC SSG-OLC-MAT GBV_ILN_70 GBV_ILN_2005 GBV_ILN_4318 54.00 AVZ AR 24 2015 1 471-483
language	English
source	Enthalten in IEEE transactions on image processing 24(2015), 1, Seite 471-483 volume:24 year:2015 number:1 pages:471-483
sourceStr	Enthalten in IEEE transactions on image processing 24(2015), 1, Seite 471-483 volume:24 year:2015 number:1 pages:471-483
format_phy_str_mv	Article
institution	findex.gbv.de
topic_facet	low-dimensional subspace learning (artificial intelligence) Covariance matrices semi-supervised splitting local maxima Estimation semi-supervised learning random forest random forest construction hyperplane separation kernel-based density estimation robust semisupervised node splitting labeled training data Training data unified optimization framework Kernel data points asymptotic mean integrated squared error criterion image classification Bandwidth computer vision Training supervised algorithms high-dimensional feature space Radio frequency subspace learning machine learning tasks quality measure margin maximization-based semisupervised methods node splitting optimisation Feature extraction Face - anatomy & histology Image Processing, Computer-Assisted - methods Pattern Recognition, Automated - methods
dewey-raw	004
isfreeaccess_bool	false
container_title	IEEE transactions on image processing
authorswithroles_txt_mv	Xiao Liu @@aut@@ Mingli Song @@oth@@ Dacheng Tao @@oth@@ Zicheng Liu @@oth@@ Luming Zhang @@oth@@ Chun Chen @@oth@@ Jiajun Bu @@oth@@
publishDateDaySort_date	2015-01-01T00:00:00Z
hierarchy_top_id	131074458
dewey-sort	14
id	OLC1959239910
language_de	englisch
fullrecord	<?xml version="1.0" encoding="UTF-8"?><collection xmlns="http://www.loc.gov/MARC21/slim"><record><leader>01000caa a2200265 4500</leader><controlfield tag="001">OLC1959239910</controlfield><controlfield tag="003">DE-627</controlfield><controlfield tag="005">20230714150846.0</controlfield><controlfield tag="007">tu</controlfield><controlfield tag="008">160206s2015 xx \|\|\|\|\| 00\| \|\|eng c</controlfield><datafield tag="024" ind1="7" ind2=" "><subfield code="a">10.1109/TIP.2014.2378017</subfield><subfield code="2">doi</subfield></datafield><datafield tag="028" ind1="5" ind2="2"><subfield code="a">PQ20160617</subfield></datafield><datafield tag="035" ind1=" " ind2=" "><subfield code="a">(DE-627)OLC1959239910</subfield></datafield><datafield tag="035" ind1=" " ind2=" "><subfield code="a">(DE-599)GBVOLC1959239910</subfield></datafield><datafield tag="035" ind1=" " ind2=" "><subfield code="a">(PRQ)c2625-33afd5198206abf0a23259fccd5db890d7a414e7e5caf12f44ca42b8d094e2120</subfield></datafield><datafield tag="035" ind1=" " ind2=" "><subfield code="a">(KEY)0213811520150000024000100471randomforestconstructionwithrobustsemisupervisedno</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="a">620</subfield><subfield code="q">DNB</subfield></datafield><datafield tag="084" ind1=" " ind2=" "><subfield code="a">54.00</subfield><subfield code="2">bkl</subfield></datafield><datafield tag="100" ind1="0" ind2=" "><subfield code="a">Xiao Liu</subfield><subfield code="e">verfasserin</subfield><subfield code="4">aut</subfield></datafield><datafield tag="245" ind1="1" ind2="0"><subfield code="a">Random Forest Construction With Robust Semisupervised Node Splitting</subfield></datafield><datafield tag="264" ind1=" " ind2="1"><subfield code="c">2015</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="520" ind1=" " ind2=" "><subfield code="a">Random forest (RF) is a very important classifier with applications in various machine learning tasks, but its promising performance heavily relies on the size of labeled training data. In this paper, we investigate constructing of RFs with a small size of labeled data and find that the performance bottleneck is located in the node splitting procedures; hence, existing solutions fail to properly partition the feature space if there are insufficient training data. To achieve robust node splitting with insufficient data, we present semisupervised splitting to overcome this limitation by splitting nodes with the guidance of both labeled and abundant unlabeled data. In particular, an accurate quality measure of node splitting is obtained by carrying out the kernel-based density estimation, whereby a multiclass version of asymptotic mean integrated squared error criterion is proposed to adaptively select the optimal bandwidth of the kernel. To avoid the curse of dimensionality, we project the data points from the original high-dimensional feature space onto a low-dimensional subspace before estimation. A unified optimization framework is proposed to select a coupled pair of subspace and separating hyperplane such that the smoothness of the subspace and the quality of the splitting are guaranteed simultaneously. Our algorithm efficiently avoids overfitting caused by bad initialization and local maxima when compared with conventional margin maximization-based semisupervised methods. We demonstrate the effectiveness of the proposed algorithm by comparing it with state-of-the-art supervised and semisupervised algorithms for typical computer vision applications, such as object categorization, face recognition, and image segmentation, on publicly available data sets.</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">low-dimensional subspace</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">learning (artificial intelligence)</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Covariance matrices</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">semi-supervised splitting</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">local maxima</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Estimation</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">semi-supervised learning</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">random forest</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">random forest construction</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">hyperplane separation</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">kernel-based density estimation</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">robust semisupervised node splitting</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">labeled training data</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Training data</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">unified optimization framework</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Kernel</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">data points</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">asymptotic mean integrated squared error criterion</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">image classification</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Bandwidth</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">computer vision</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Training</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">supervised algorithms</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">high-dimensional feature space</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Radio frequency</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">subspace learning</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">machine learning tasks</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">quality measure</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">margin maximization-based semisupervised methods</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">node splitting</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">optimisation</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Feature extraction</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Face - anatomy & histology</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Image Processing, Computer-Assisted - methods</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Pattern Recognition, Automated - methods</subfield></datafield><datafield tag="700" ind1="0" ind2=" "><subfield code="a">Mingli Song</subfield><subfield code="4">oth</subfield></datafield><datafield tag="700" ind1="0" ind2=" "><subfield code="a">Dacheng Tao</subfield><subfield code="4">oth</subfield></datafield><datafield tag="700" ind1="0" ind2=" "><subfield code="a">Zicheng Liu</subfield><subfield code="4">oth</subfield></datafield><datafield tag="700" ind1="0" ind2=" "><subfield code="a">Luming Zhang</subfield><subfield code="4">oth</subfield></datafield><datafield tag="700" ind1="0" ind2=" "><subfield code="a">Chun Chen</subfield><subfield code="4">oth</subfield></datafield><datafield tag="700" ind1="0" ind2=" "><subfield code="a">Jiajun Bu</subfield><subfield code="4">oth</subfield></datafield><datafield tag="773" ind1="0" ind2="8"><subfield code="i">Enthalten in</subfield><subfield code="t">IEEE transactions on image processing</subfield><subfield code="d">New York, NY : Inst., 1992</subfield><subfield code="g">24(2015), 1, Seite 471-483</subfield><subfield code="w">(DE-627)131074458</subfield><subfield code="w">(DE-600)1111265-7</subfield><subfield code="w">(DE-576)029165008</subfield><subfield code="x">1057-7149</subfield><subfield code="7">nnns</subfield></datafield><datafield tag="773" ind1="1" ind2="8"><subfield code="g">volume:24</subfield><subfield code="g">year:2015</subfield><subfield code="g">number:1</subfield><subfield code="g">pages:471-483</subfield></datafield><datafield tag="856" ind1="4" ind2="1"><subfield code="u">http://dx.doi.org/10.1109/TIP.2014.2378017</subfield><subfield code="3">Volltext</subfield></datafield><datafield tag="856" ind1="4" ind2="2"><subfield code="u">http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=6975199</subfield></datafield><datafield tag="856" ind1="4" ind2="2"><subfield code="u">http://www.ncbi.nlm.nih.gov/pubmed/25494503</subfield></datafield><datafield tag="856" ind1="4" ind2="2"><subfield code="u">http://search.proquest.com/docview/1640798292</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-TEC</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_2005</subfield></datafield><datafield tag="912" ind1=" " ind2=" "><subfield code="a">GBV_ILN_4318</subfield></datafield><datafield tag="936" ind1="b" ind2="k"><subfield code="a">54.00</subfield><subfield code="q">AVZ</subfield></datafield><datafield tag="951" ind1=" " ind2=" "><subfield code="a">AR</subfield></datafield><datafield tag="952" ind1=" " ind2=" "><subfield code="d">24</subfield><subfield code="j">2015</subfield><subfield code="e">1</subfield><subfield code="h">471-483</subfield></datafield></record></collection>
author	Xiao Liu
spellingShingle	Xiao Liu ddc 004 bkl 54.00 misc low-dimensional subspace misc learning (artificial intelligence) misc Covariance matrices misc semi-supervised splitting misc local maxima misc Estimation misc semi-supervised learning misc random forest misc random forest construction misc hyperplane separation misc kernel-based density estimation misc robust semisupervised node splitting misc labeled training data misc Training data misc unified optimization framework misc Kernel misc data points misc asymptotic mean integrated squared error criterion misc image classification misc Bandwidth misc computer vision misc Training misc supervised algorithms misc high-dimensional feature space misc Radio frequency misc subspace learning misc machine learning tasks misc quality measure misc margin maximization-based semisupervised methods misc node splitting misc optimisation misc Feature extraction misc Face - anatomy & histology misc Image Processing, Computer-Assisted - methods misc Pattern Recognition, Automated - methods Random Forest Construction With Robust Semisupervised Node Splitting
authorStr	Xiao Liu
ppnlink_with_tag_str_mv	@@773@@(DE-627)131074458
format	Article
dewey-ones	004 - Data processing & computer science 620 - Engineering & allied operations
delete_txt_mv	keep
author_role	aut
collection	OLC
remote_str	false
illustrated	Not Illustrated
issn	1057-7149
topic_title	004 620 DNB 54.00 bkl Random Forest Construction With Robust Semisupervised Node Splitting low-dimensional subspace learning (artificial intelligence) Covariance matrices semi-supervised splitting local maxima Estimation semi-supervised learning random forest random forest construction hyperplane separation kernel-based density estimation robust semisupervised node splitting labeled training data Training data unified optimization framework Kernel data points asymptotic mean integrated squared error criterion image classification Bandwidth computer vision Training supervised algorithms high-dimensional feature space Radio frequency subspace learning machine learning tasks quality measure margin maximization-based semisupervised methods node splitting optimisation Feature extraction Face - anatomy & histology Image Processing, Computer-Assisted - methods Pattern Recognition, Automated - methods
topic	ddc 004 bkl 54.00 misc low-dimensional subspace misc learning (artificial intelligence) misc Covariance matrices misc semi-supervised splitting misc local maxima misc Estimation misc semi-supervised learning misc random forest misc random forest construction misc hyperplane separation misc kernel-based density estimation misc robust semisupervised node splitting misc labeled training data misc Training data misc unified optimization framework misc Kernel misc data points misc asymptotic mean integrated squared error criterion misc image classification misc Bandwidth misc computer vision misc Training misc supervised algorithms misc high-dimensional feature space misc Radio frequency misc subspace learning misc machine learning tasks misc quality measure misc margin maximization-based semisupervised methods misc node splitting misc optimisation misc Feature extraction misc Face - anatomy & histology misc Image Processing, Computer-Assisted - methods misc Pattern Recognition, Automated - methods
topic_unstemmed	ddc 004 bkl 54.00 misc low-dimensional subspace misc learning (artificial intelligence) misc Covariance matrices misc semi-supervised splitting misc local maxima misc Estimation misc semi-supervised learning misc random forest misc random forest construction misc hyperplane separation misc kernel-based density estimation misc robust semisupervised node splitting misc labeled training data misc Training data misc unified optimization framework misc Kernel misc data points misc asymptotic mean integrated squared error criterion misc image classification misc Bandwidth misc computer vision misc Training misc supervised algorithms misc high-dimensional feature space misc Radio frequency misc subspace learning misc machine learning tasks misc quality measure misc margin maximization-based semisupervised methods misc node splitting misc optimisation misc Feature extraction misc Face - anatomy & histology misc Image Processing, Computer-Assisted - methods misc Pattern Recognition, Automated - methods
topic_browse	ddc 004 bkl 54.00 misc low-dimensional subspace misc learning (artificial intelligence) misc Covariance matrices misc semi-supervised splitting misc local maxima misc Estimation misc semi-supervised learning misc random forest misc random forest construction misc hyperplane separation misc kernel-based density estimation misc robust semisupervised node splitting misc labeled training data misc Training data misc unified optimization framework misc Kernel misc data points misc asymptotic mean integrated squared error criterion misc image classification misc Bandwidth misc computer vision misc Training misc supervised algorithms misc high-dimensional feature space misc Radio frequency misc subspace learning misc machine learning tasks misc quality measure misc margin maximization-based semisupervised methods misc node splitting misc optimisation misc Feature extraction misc Face - anatomy & histology misc Image Processing, Computer-Assisted - methods misc Pattern Recognition, Automated - methods
format_facet	Aufsätze Gedruckte Aufsätze
format_main_str_mv	Text Zeitschrift/Artikel
carriertype_str_mv	nc
author2_variant	m s ms d t dt z l zl l z lz c c cc j b jb
hierarchy_parent_title	IEEE transactions on image processing
hierarchy_parent_id	131074458
dewey-tens	000 - Computer science, knowledge & systems 620 - Engineering
hierarchy_top_title	IEEE transactions on image processing
isfreeaccess_txt	false
familylinks_str_mv	(DE-627)131074458 (DE-600)1111265-7 (DE-576)029165008
title	Random Forest Construction With Robust Semisupervised Node Splitting
ctrlnum	(DE-627)OLC1959239910 (DE-599)GBVOLC1959239910 (PRQ)c2625-33afd5198206abf0a23259fccd5db890d7a414e7e5caf12f44ca42b8d094e2120 (KEY)0213811520150000024000100471randomforestconstructionwithrobustsemisupervisedno
title_full	Random Forest Construction With Robust Semisupervised Node Splitting
author_sort	Xiao Liu
journal	IEEE transactions on image processing
journalStr	IEEE transactions on image processing
lang_code	eng
isOA_bool	false
dewey-hundreds	000 - Computer science, information & general works 600 - Technology
recordtype	marc
publishDateSort	2015
contenttype_str_mv	txt
container_start_page	471
author_browse	Xiao Liu
container_volume	24
class	004 620 DNB 54.00 bkl
format_se	Aufsätze
author-letter	Xiao Liu
doi_str_mv	10.1109/TIP.2014.2378017
dewey-full	004 620
title_sort	random forest construction with robust semisupervised node splitting
title_auth	Random Forest Construction With Robust Semisupervised Node Splitting
abstract	Random forest (RF) is a very important classifier with applications in various machine learning tasks, but its promising performance heavily relies on the size of labeled training data. In this paper, we investigate constructing of RFs with a small size of labeled data and find that the performance bottleneck is located in the node splitting procedures; hence, existing solutions fail to properly partition the feature space if there are insufficient training data. To achieve robust node splitting with insufficient data, we present semisupervised splitting to overcome this limitation by splitting nodes with the guidance of both labeled and abundant unlabeled data. In particular, an accurate quality measure of node splitting is obtained by carrying out the kernel-based density estimation, whereby a multiclass version of asymptotic mean integrated squared error criterion is proposed to adaptively select the optimal bandwidth of the kernel. To avoid the curse of dimensionality, we project the data points from the original high-dimensional feature space onto a low-dimensional subspace before estimation. A unified optimization framework is proposed to select a coupled pair of subspace and separating hyperplane such that the smoothness of the subspace and the quality of the splitting are guaranteed simultaneously. Our algorithm efficiently avoids overfitting caused by bad initialization and local maxima when compared with conventional margin maximization-based semisupervised methods. We demonstrate the effectiveness of the proposed algorithm by comparing it with state-of-the-art supervised and semisupervised algorithms for typical computer vision applications, such as object categorization, face recognition, and image segmentation, on publicly available data sets.
abstractGer	Random forest (RF) is a very important classifier with applications in various machine learning tasks, but its promising performance heavily relies on the size of labeled training data. In this paper, we investigate constructing of RFs with a small size of labeled data and find that the performance bottleneck is located in the node splitting procedures; hence, existing solutions fail to properly partition the feature space if there are insufficient training data. To achieve robust node splitting with insufficient data, we present semisupervised splitting to overcome this limitation by splitting nodes with the guidance of both labeled and abundant unlabeled data. In particular, an accurate quality measure of node splitting is obtained by carrying out the kernel-based density estimation, whereby a multiclass version of asymptotic mean integrated squared error criterion is proposed to adaptively select the optimal bandwidth of the kernel. To avoid the curse of dimensionality, we project the data points from the original high-dimensional feature space onto a low-dimensional subspace before estimation. A unified optimization framework is proposed to select a coupled pair of subspace and separating hyperplane such that the smoothness of the subspace and the quality of the splitting are guaranteed simultaneously. Our algorithm efficiently avoids overfitting caused by bad initialization and local maxima when compared with conventional margin maximization-based semisupervised methods. We demonstrate the effectiveness of the proposed algorithm by comparing it with state-of-the-art supervised and semisupervised algorithms for typical computer vision applications, such as object categorization, face recognition, and image segmentation, on publicly available data sets.
abstract_unstemmed	Random forest (RF) is a very important classifier with applications in various machine learning tasks, but its promising performance heavily relies on the size of labeled training data. In this paper, we investigate constructing of RFs with a small size of labeled data and find that the performance bottleneck is located in the node splitting procedures; hence, existing solutions fail to properly partition the feature space if there are insufficient training data. To achieve robust node splitting with insufficient data, we present semisupervised splitting to overcome this limitation by splitting nodes with the guidance of both labeled and abundant unlabeled data. In particular, an accurate quality measure of node splitting is obtained by carrying out the kernel-based density estimation, whereby a multiclass version of asymptotic mean integrated squared error criterion is proposed to adaptively select the optimal bandwidth of the kernel. To avoid the curse of dimensionality, we project the data points from the original high-dimensional feature space onto a low-dimensional subspace before estimation. A unified optimization framework is proposed to select a coupled pair of subspace and separating hyperplane such that the smoothness of the subspace and the quality of the splitting are guaranteed simultaneously. Our algorithm efficiently avoids overfitting caused by bad initialization and local maxima when compared with conventional margin maximization-based semisupervised methods. We demonstrate the effectiveness of the proposed algorithm by comparing it with state-of-the-art supervised and semisupervised algorithms for typical computer vision applications, such as object categorization, face recognition, and image segmentation, on publicly available data sets.
collection_details	GBV_USEFLAG_A SYSFLAG_A GBV_OLC SSG-OLC-TEC SSG-OLC-MAT GBV_ILN_70 GBV_ILN_2005 GBV_ILN_4318
container_issue	1
title_short	Random Forest Construction With Robust Semisupervised Node Splitting
url	http://dx.doi.org/10.1109/TIP.2014.2378017 http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=6975199 http://www.ncbi.nlm.nih.gov/pubmed/25494503 http://search.proquest.com/docview/1640798292
remote_bool	false
author2	Mingli Song Dacheng Tao Zicheng Liu Luming Zhang Chun Chen Jiajun Bu
author2Str	Mingli Song Dacheng Tao Zicheng Liu Luming Zhang Chun Chen Jiajun Bu
ppnlink	131074458
mediatype_str_mv	n
isOA_txt	false
hochschulschrift_bool	false
author2_role	oth oth oth oth oth oth
doi_str	10.1109/TIP.2014.2378017
up_date	2024-07-03T16:25:35.765Z
_version_	1803575818469769216
fullrecord_marcxml	<?xml version="1.0" encoding="UTF-8"?><collection xmlns="http://www.loc.gov/MARC21/slim"><record><leader>01000caa a2200265 4500</leader><controlfield tag="001">OLC1959239910</controlfield><controlfield tag="003">DE-627</controlfield><controlfield tag="005">20230714150846.0</controlfield><controlfield tag="007">tu</controlfield><controlfield tag="008">160206s2015 xx \|\|\|\|\| 00\| \|\|eng c</controlfield><datafield tag="024" ind1="7" ind2=" "><subfield code="a">10.1109/TIP.2014.2378017</subfield><subfield code="2">doi</subfield></datafield><datafield tag="028" ind1="5" ind2="2"><subfield code="a">PQ20160617</subfield></datafield><datafield tag="035" ind1=" " ind2=" "><subfield code="a">(DE-627)OLC1959239910</subfield></datafield><datafield tag="035" ind1=" " ind2=" "><subfield code="a">(DE-599)GBVOLC1959239910</subfield></datafield><datafield tag="035" ind1=" " ind2=" "><subfield code="a">(PRQ)c2625-33afd5198206abf0a23259fccd5db890d7a414e7e5caf12f44ca42b8d094e2120</subfield></datafield><datafield tag="035" ind1=" " ind2=" "><subfield code="a">(KEY)0213811520150000024000100471randomforestconstructionwithrobustsemisupervisedno</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="a">620</subfield><subfield code="q">DNB</subfield></datafield><datafield tag="084" ind1=" " ind2=" "><subfield code="a">54.00</subfield><subfield code="2">bkl</subfield></datafield><datafield tag="100" ind1="0" ind2=" "><subfield code="a">Xiao Liu</subfield><subfield code="e">verfasserin</subfield><subfield code="4">aut</subfield></datafield><datafield tag="245" ind1="1" ind2="0"><subfield code="a">Random Forest Construction With Robust Semisupervised Node Splitting</subfield></datafield><datafield tag="264" ind1=" " ind2="1"><subfield code="c">2015</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="520" ind1=" " ind2=" "><subfield code="a">Random forest (RF) is a very important classifier with applications in various machine learning tasks, but its promising performance heavily relies on the size of labeled training data. In this paper, we investigate constructing of RFs with a small size of labeled data and find that the performance bottleneck is located in the node splitting procedures; hence, existing solutions fail to properly partition the feature space if there are insufficient training data. To achieve robust node splitting with insufficient data, we present semisupervised splitting to overcome this limitation by splitting nodes with the guidance of both labeled and abundant unlabeled data. In particular, an accurate quality measure of node splitting is obtained by carrying out the kernel-based density estimation, whereby a multiclass version of asymptotic mean integrated squared error criterion is proposed to adaptively select the optimal bandwidth of the kernel. To avoid the curse of dimensionality, we project the data points from the original high-dimensional feature space onto a low-dimensional subspace before estimation. A unified optimization framework is proposed to select a coupled pair of subspace and separating hyperplane such that the smoothness of the subspace and the quality of the splitting are guaranteed simultaneously. Our algorithm efficiently avoids overfitting caused by bad initialization and local maxima when compared with conventional margin maximization-based semisupervised methods. We demonstrate the effectiveness of the proposed algorithm by comparing it with state-of-the-art supervised and semisupervised algorithms for typical computer vision applications, such as object categorization, face recognition, and image segmentation, on publicly available data sets.</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">low-dimensional subspace</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">learning (artificial intelligence)</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Covariance matrices</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">semi-supervised splitting</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">local maxima</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Estimation</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">semi-supervised learning</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">random forest</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">random forest construction</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">hyperplane separation</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">kernel-based density estimation</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">robust semisupervised node splitting</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">labeled training data</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Training data</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">unified optimization framework</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Kernel</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">data points</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">asymptotic mean integrated squared error criterion</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">image classification</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Bandwidth</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">computer vision</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Training</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">supervised algorithms</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">high-dimensional feature space</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Radio frequency</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">subspace learning</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">machine learning tasks</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">quality measure</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">margin maximization-based semisupervised methods</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">node splitting</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">optimisation</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Feature extraction</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Face - anatomy & histology</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Image Processing, Computer-Assisted - methods</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Pattern Recognition, Automated - methods</subfield></datafield><datafield tag="700" ind1="0" ind2=" "><subfield code="a">Mingli Song</subfield><subfield code="4">oth</subfield></datafield><datafield tag="700" ind1="0" ind2=" "><subfield code="a">Dacheng Tao</subfield><subfield code="4">oth</subfield></datafield><datafield tag="700" ind1="0" ind2=" "><subfield code="a">Zicheng Liu</subfield><subfield code="4">oth</subfield></datafield><datafield tag="700" ind1="0" ind2=" "><subfield code="a">Luming Zhang</subfield><subfield code="4">oth</subfield></datafield><datafield tag="700" ind1="0" ind2=" "><subfield code="a">Chun Chen</subfield><subfield code="4">oth</subfield></datafield><datafield tag="700" ind1="0" ind2=" "><subfield code="a">Jiajun Bu</subfield><subfield code="4">oth</subfield></datafield><datafield tag="773" ind1="0" ind2="8"><subfield code="i">Enthalten in</subfield><subfield code="t">IEEE transactions on image processing</subfield><subfield code="d">New York, NY : Inst., 1992</subfield><subfield code="g">24(2015), 1, Seite 471-483</subfield><subfield code="w">(DE-627)131074458</subfield><subfield code="w">(DE-600)1111265-7</subfield><subfield code="w">(DE-576)029165008</subfield><subfield code="x">1057-7149</subfield><subfield code="7">nnns</subfield></datafield><datafield tag="773" ind1="1" ind2="8"><subfield code="g">volume:24</subfield><subfield code="g">year:2015</subfield><subfield code="g">number:1</subfield><subfield code="g">pages:471-483</subfield></datafield><datafield tag="856" ind1="4" ind2="1"><subfield code="u">http://dx.doi.org/10.1109/TIP.2014.2378017</subfield><subfield code="3">Volltext</subfield></datafield><datafield tag="856" ind1="4" ind2="2"><subfield code="u">http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=6975199</subfield></datafield><datafield tag="856" ind1="4" ind2="2"><subfield code="u">http://www.ncbi.nlm.nih.gov/pubmed/25494503</subfield></datafield><datafield tag="856" ind1="4" ind2="2"><subfield code="u">http://search.proquest.com/docview/1640798292</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-TEC</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_2005</subfield></datafield><datafield tag="912" ind1=" " ind2=" "><subfield code="a">GBV_ILN_4318</subfield></datafield><datafield tag="936" ind1="b" ind2="k"><subfield code="a">54.00</subfield><subfield code="q">AVZ</subfield></datafield><datafield tag="951" ind1=" " ind2=" "><subfield code="a">AR</subfield></datafield><datafield tag="952" ind1=" " ind2=" "><subfield code="d">24</subfield><subfield code="j">2015</subfield><subfield code="e">1</subfield><subfield code="h">471-483</subfield></datafield></record></collection>
score	7.4008274

Nicht das Richtige dabei?

Schreiben Sie uns!

Random Forest Construction With Robust Semisupervised Node Splitting

Nicht das Richtige dabei?

Zugang & Verfügbarkeit

Vorhandene Bände

Nicht das Richtige dabei?