Pub Date : 2004-08-23DOI: 10.1109/ICPR.2004.1334540
P. Adeodato, G. C. Vasconcelos, A. L. Arnaud, R. A. F. Santos, Rodrigo C. L. V. Cunha, Domingos S. M. P. Monteiro
Neural networks and logistic regression have been among the most widely used AI technique in applications of pattern classification.Much has been discussed about if there is any significant difference in between them but much less has been actually done with real-world applications data (large scale) to help settle this matter, with a few exceptions.This paper presents a performance comparison between these two techniques on the market application of credit risk assessment, making use of a large database from an outstanding credit bureau and financial institution (a sample of 180,000 examples).The comparison was carried out through a 30-fold stratified cross-validation process to define the confidence intervals for the performance evaluation. Several metrics were applied both on the optimal decision point and along the continuous output domain.The statistical tests showed that multilayer perceptrons perform better than logistic regression at 95% confidence level, for all the metrics used.
{"title":"Neural Networks vs Logistic Regression: a Comparative Study on a Large Data Set","authors":"P. Adeodato, G. C. Vasconcelos, A. L. Arnaud, R. A. F. Santos, Rodrigo C. L. V. Cunha, Domingos S. M. P. Monteiro","doi":"10.1109/ICPR.2004.1334540","DOIUrl":"https://doi.org/10.1109/ICPR.2004.1334540","url":null,"abstract":"Neural networks and logistic regression have been among the most widely used AI technique in applications of pattern classification.Much has been discussed about if there is any significant difference in between them but much less has been actually done with real-world applications data (large scale) to help settle this matter, with a few exceptions.This paper presents a performance comparison between these two techniques on the market application of credit risk assessment, making use of a large database from an outstanding credit bureau and financial institution (a sample of 180,000 examples).The comparison was carried out through a 30-fold stratified cross-validation process to define the confidence intervals for the performance evaluation. Several metrics were applied both on the optimal decision point and along the continuous output domain.The statistical tests showed that multilayer perceptrons perform better than logistic regression at 95% confidence level, for all the metrics used.","PeriodicalId":74516,"journal":{"name":"Proceedings of the ... IAPR International Conference on Pattern Recognition. International Conference on Pattern Recognition","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2004-08-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"73831228","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2004-08-23DOI: 10.1109/ICPR.2004.1334454
M. Land
Summary form only given, as follows. Humans, with the massive computational power of the cerebral cortex, have managed to solve most of the problems that make pattern recognition such a difficult task. Other animals are not so well endowed with processing power: an insect brain, for example, has 105 to 106 neurons compared with our 1011. Nevertheless, they still have to recognise predators, prey and conspecifics, and find their way around the world. Often this means that they have to cut corners, using what machinery they have in economical ways. Typically this means tailoring their recognition systems to just those features that really matter, rather than going for the general purpose mechanism that primates have achieved. In this talk I will examine some of the ingenious and sometimes strange solutions that animals such as insects, spiders, crabs and molluscs have come up with to simplify the tasks of pattern recognition, while still satisfying their requirements of their often complex behaviour.
{"title":"Pattern Perception in Animals Remote from Man","authors":"M. Land","doi":"10.1109/ICPR.2004.1334454","DOIUrl":"https://doi.org/10.1109/ICPR.2004.1334454","url":null,"abstract":"Summary form only given, as follows. Humans, with the massive computational power of the cerebral cortex, have managed to solve most of the problems that make pattern recognition such a difficult task. Other animals are not so well endowed with processing power: an insect brain, for example, has 105 to 106 neurons compared with our 1011. Nevertheless, they still have to recognise predators, prey and conspecifics, and find their way around the world. Often this means that they have to cut corners, using what machinery they have in economical ways. Typically this means tailoring their recognition systems to just those features that really matter, rather than going for the general purpose mechanism that primates have achieved. In this talk I will examine some of the ingenious and sometimes strange solutions that animals such as insects, spiders, crabs and molluscs have come up with to simplify the tasks of pattern recognition, while still satisfying their requirements of their often complex behaviour.","PeriodicalId":74516,"journal":{"name":"Proceedings of the ... IAPR International Conference on Pattern Recognition. International Conference on Pattern Recognition","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2004-08-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"76554587","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2004-08-23DOI: 10.1109/ICPR.2004.1334096
C. Wöhler
A novel framework for three-dimensional surface reconstruction by self-consistent fusion of shading and shadow features is presented. Based on the analysis of at least two pixel-synchronous images of the scene under different illumination conditions, this framework combines a shape from shading approach for estimating surface gradients and altitude variations with a shadow analysis that allows for an accurate determination of altitude differences on the surface. As a first step, the result of shadow analysis is used for selecting a consistent solution of the shape from shading reconstruction algorithm. As a second step, an additional error term derived from the fine structure of the shadow is incorporated into the reconstruction algorithm. This framework is applied to three-dimensional reconstruction of regions on the lunar surface using ground based CCD images. Beyond the planetary science scenario, it is applicable to classical machine vision tasks such as surface inspection in the context of industrial quality control.
{"title":"3D Surface Reconstruction by Self-Consistent Fusion of Shading and Shadow Features","authors":"C. Wöhler","doi":"10.1109/ICPR.2004.1334096","DOIUrl":"https://doi.org/10.1109/ICPR.2004.1334096","url":null,"abstract":"A novel framework for three-dimensional surface reconstruction by self-consistent fusion of shading and shadow features is presented. Based on the analysis of at least two pixel-synchronous images of the scene under different illumination conditions, this framework combines a shape from shading approach for estimating surface gradients and altitude variations with a shadow analysis that allows for an accurate determination of altitude differences on the surface. As a first step, the result of shadow analysis is used for selecting a consistent solution of the shape from shading reconstruction algorithm. As a second step, an additional error term derived from the fine structure of the shadow is incorporated into the reconstruction algorithm. This framework is applied to three-dimensional reconstruction of regions on the lunar surface using ground based CCD images. Beyond the planetary science scenario, it is applicable to classical machine vision tasks such as surface inspection in the context of industrial quality control.","PeriodicalId":74516,"journal":{"name":"Proceedings of the ... IAPR International Conference on Pattern Recognition. International Conference on Pattern Recognition","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2004-08-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"78641368","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2004-08-23DOI: 10.1109/ICPR.2004.1334558
Ville Hautamäki, Ismo Kärkkäinen, P. Fränti
We present an outlier detection using indegree number (ODIN) algorithm that utilizes k-nearest neighbour graph. Improvements to existing kNN distance-based method are also proposed. We compare the methods with real and synthetic datasets. The results show that the proposed method achieves reasonable results with synthetic data and outperforms compared methods with real data sets with small number of observations.
{"title":"Outlier Detection Using k-Nearest Neighbour Graph","authors":"Ville Hautamäki, Ismo Kärkkäinen, P. Fränti","doi":"10.1109/ICPR.2004.1334558","DOIUrl":"https://doi.org/10.1109/ICPR.2004.1334558","url":null,"abstract":"We present an outlier detection using indegree number (ODIN) algorithm that utilizes k-nearest neighbour graph. Improvements to existing kNN distance-based method are also proposed. We compare the methods with real and synthetic datasets. The results show that the proposed method achieves reasonable results with synthetic data and outperforms compared methods with real data sets with small number of observations.","PeriodicalId":74516,"journal":{"name":"Proceedings of the ... IAPR International Conference on Pattern Recognition. International Conference on Pattern Recognition","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2004-08-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"88244736","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2004-01-01DOI: 10.1109/ICPR.2004.1334061
D. Kropotov, D. Vetrov
Although using fuzzy logic in control systems has become widely established as an appropriate approach, its application in area of pattern recognition and data mining seems to be restricted. These systems have several bottlenecks mainly concerning fuzzy rules generation and fuzzy sets forming. The state-of-the-art technique here is neuro-fuzzy approach which has some disadvantages. In the presented article there considered an algorithm for rules generation based on alternative principles.
{"title":"An Algorithm for Rule Generation in Fuzzy Expert Systems","authors":"D. Kropotov, D. Vetrov","doi":"10.1109/ICPR.2004.1334061","DOIUrl":"https://doi.org/10.1109/ICPR.2004.1334061","url":null,"abstract":"Although using fuzzy logic in control systems has become widely established as an appropriate approach, its application in area of pattern recognition and data mining seems to be restricted. These systems have several bottlenecks mainly concerning fuzzy rules generation and fuzzy sets forming. The state-of-the-art technique here is neuro-fuzzy approach which has some disadvantages. In the presented article there considered an algorithm for rules generation based on alternative principles.","PeriodicalId":74516,"journal":{"name":"Proceedings of the ... IAPR International Conference on Pattern Recognition. International Conference on Pattern Recognition","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2004-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"87287532","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2004-01-01DOI: 10.1109/ICPR.2004.1334387
Hongliang Bai, Chang-ping Liu
{"title":"A hybrid License Plate Extraction Method Based On Edge Statistics and Morphology","authors":"Hongliang Bai, Chang-ping Liu","doi":"10.1109/ICPR.2004.1334387","DOIUrl":"https://doi.org/10.1109/ICPR.2004.1334387","url":null,"abstract":"","PeriodicalId":74516,"journal":{"name":"Proceedings of the ... IAPR International Conference on Pattern Recognition. International Conference on Pattern Recognition","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2004-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"77839702","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2004-01-01DOI: 10.1109/ICPR.2004.1334511
Jiaoyan Ai, Di Liu, Xuefeng Zhu
{"title":"Combination of Wavelet Analysis and Color Applied to Automatic Color Grading of Ceramic Tiles","authors":"Jiaoyan Ai, Di Liu, Xuefeng Zhu","doi":"10.1109/ICPR.2004.1334511","DOIUrl":"https://doi.org/10.1109/ICPR.2004.1334511","url":null,"abstract":"","PeriodicalId":74516,"journal":{"name":"Proceedings of the ... IAPR International Conference on Pattern Recognition. International Conference on Pattern Recognition","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2004-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"89335955","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2004-01-01DOI: 10.1109/ICPR.2004.1333792
R. A. Mollineda
An iterative learning method to update labeled string prototypes for a 1-nearest prototype (1-np) classification is introduced. Given a (typically reduced) set of initial string prototypes and a training set, it iteratively updates prototypes to better discriminate training samples. The update rule, which is based on the edit distance, adjusts a prototype by removing those local differences which are both frequent with respect to same-class closer training strings and infrequent with respect to different-class closer training strings. Closer training strings are defined by unsupervised clustering. The process continues until prototypes converge. Its main innovation is to provide a non-random local update rule to “move” a string prototype towards a number of string samples. A series of learning/classification experiments show a better 1-np performance of the updated prototypes with respect to the initial ones, that were originally selected to guarantee a good classification.
{"title":"A Learning Model for Multiple-Prototype Classification of Strings","authors":"R. A. Mollineda","doi":"10.1109/ICPR.2004.1333792","DOIUrl":"https://doi.org/10.1109/ICPR.2004.1333792","url":null,"abstract":"An iterative learning method to update labeled string prototypes for a 1-nearest prototype (1-np) classification is introduced. Given a (typically reduced) set of initial string prototypes and a training set, it iteratively updates prototypes to better discriminate training samples. The update rule, which is based on the edit distance, adjusts a prototype by removing those local differences which are both frequent with respect to same-class closer training strings and infrequent with respect to different-class closer training strings. Closer training strings are defined by unsupervised clustering. The process continues until prototypes converge. Its main innovation is to provide a non-random local update rule to “move” a string prototype towards a number of string samples. A series of learning/classification experiments show a better 1-np performance of the updated prototypes with respect to the initial ones, that were originally selected to guarantee a good classification.","PeriodicalId":74516,"journal":{"name":"Proceedings of the ... IAPR International Conference on Pattern Recognition. International Conference on Pattern Recognition","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2004-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"85284829","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2004-01-01DOI: 10.1109/ICPR.2004.1334166
Y. Belaïd, A. Belaïd
In this paper a morphological tagging approach for document image invoice analysis is described. Tokens close by their morphology and confirmed in their location within different similar contexts make apparent some parts of speech representative of the structure elements. This bottom up approach avoids the use of an priori knowledge provided that there are redundant and frequent contexts in the text. The approach is applied on the invoice body text roughly recognized by OCR and automatically segmented. The method makes possible the detection of the invoice articles and their different fields. The regularity of the article composition and its redundancy in the invoice is a good help for its structure. The recognition rate of 276 invoices and 1704 articles, is over than 91.02% for articles and 92.56% for fields.
{"title":"Morphological Tagging Approach in Document Analysis of Invoices","authors":"Y. Belaïd, A. Belaïd","doi":"10.1109/ICPR.2004.1334166","DOIUrl":"https://doi.org/10.1109/ICPR.2004.1334166","url":null,"abstract":"In this paper a morphological tagging approach for document image invoice analysis is described. Tokens close by their morphology and confirmed in their location within different similar contexts make apparent some parts of speech representative of the structure elements. This bottom up approach avoids the use of an priori knowledge provided that there are redundant and frequent contexts in the text. The approach is applied on the invoice body text roughly recognized by OCR and automatically segmented. The method makes possible the detection of the invoice articles and their different fields. The regularity of the article composition and its redundancy in the invoice is a good help for its structure. The recognition rate of 276 invoices and 1704 articles, is over than 91.02% for articles and 92.56% for fields.","PeriodicalId":74516,"journal":{"name":"Proceedings of the ... IAPR International Conference on Pattern Recognition. International Conference on Pattern Recognition","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2004-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"91415768","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2004-01-01DOI: 10.1109/ICPR.2004.1334600
Junwen Wu, Mohan M. Trivedi, Bhaskar D. Rao
{"title":"High Frequency Component Compensation based Super-Resolution Algorithm for Face Video Enhancement","authors":"Junwen Wu, Mohan M. Trivedi, Bhaskar D. Rao","doi":"10.1109/ICPR.2004.1334600","DOIUrl":"https://doi.org/10.1109/ICPR.2004.1334600","url":null,"abstract":"","PeriodicalId":74516,"journal":{"name":"Proceedings of the ... IAPR International Conference on Pattern Recognition. International Conference on Pattern Recognition","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2004-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"75074692","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：