1,720,995 research outputs found
Using Co-training and Self-training in Semi-Supervised Multiple Classifier Systems
No full textMultiple classifier systems have been originally proposed for supervised classification tasks, and few works have dealt with semi-supervised multiple classifiers. However, there are important pattern recognition applications, such as multi-sensor remote sensing and multi-modal biometrics, which demand semi-supervised multiple classifier systems able to exploit both labelled and unlabelled data. In this paper, the use, in multiple classifier systems, of two well known semi-supervised learning methods, namely, co-training and self-training, is investigated by experiments. Reported results on benchmarking data sets show that co-training and self-training allow exploiting unlabelled data in different types of multiple classifiers systems
Template co-update in multimodal biometric systems
No full textPerformances of biometric recognition systems can degrade quickly when the input biometric traits exhibit substantial variations compared to the templates collected during the enrolment stage of users. This issue can be addressed using template update methods. In this paper, a novel template update method based on the concept of biometric co-training is presented. In multimodal biometric systems, this method allows co-updating the template galleries of different biometrics, realizing a co-training process of biometric experts which allows updating templates more quickly and effectively. Reported results provide a first experimental evidence of the effectiveness of the proposed template update method
Ensemble learning for Intrusion Detection in Computer Networks
No full textThe security of computer networks plays a strategic role in modern
computer systems. In order to enforce high protection levels against threats, a
number of software tools are currently developed. Intrusion Detection Systems aim
at detecting intruder who eluded the "first line" protection. In this paper, a pattern
recognition approach to network intrusion detection based on ensemble learning
paradigms is proposed. The potentialities of such an approach for data fusion and
some open issues are outline
Personal identity verification by serial fusion of fingerprint and face matchers
No full textThe use of personal identity verification systems with multi-modal biometrics has been proposed in order to increase the performance and robustness against environmental variations and fraudulent attacks. Usually multi-modal fusion of biometrics is performed in parallel at the score-level by combining the individual matching scores. This parallel strategy exhibits some drawbacks: (i) all available biometrics are necessary to perform fusion, thus the verification time depends on the slowest system; (ii) some users could be easily recognizable using a certain biometric instead of another one and (iii) the system invasiveness increases. A system characterized by the serial combination of multiple biometrics can be a good trade-off between verification time, performance and acceptability. However, these systems have been poorly investigated, and no support for designing the processing chain has been given so far. In this paper, we propose a novel serial scheme and a simple mathematical model able to predict the performance of two serially combined matchers as function of the selected processing chain. Our model helps the designer in finding the processing chain allowing a trade-off, in particular, between performance and matching time. Experiments carried out on well-known benchmark data sets made up of face and fingerprint images support the usefulness of the proposed methodology and compare it with standard parallel fusion
Semi-supervised techniques for improving the performance of multiple classifier systems and personal recognition systems using biometric traits
No full textIn this paper, some research activities carried out by the Pattern Recognition and Applications Group of the University of Cagliari are presented. From the methodological viewpoint, the research activities of the group have been focused since its foundation on the field of pattern recognition. In this paper, we describe some research activities with particular reference to the development of semi-supervised algorithms applied to multiple classifier systems. From the viewpoint of applied research, the group is currently involved in the development of semi-supervised techniques for security applications. In this paper, we focus on the personal authentication/identification through biometrics. These research activities are currently funded by several contracts and grants from private firms and government agencies
Multimodal fingerprint verification by score-level fusion: an experimental investigation
No full textThe score-level fusion approaches for fingerprint verification have been widely investigated. However, this investigation
has been performed by studying each approach independently from the others, thus using different acquisition sensors, matching
algorithms, fusion rules, and data sets. Due to this strong variability, the literature is lack of an experimental investigation aimed
to fairly compare the various approaches. This is the scope of the present paper, from the point of view of the performance
improvement especially. In our opinion, this investigation can allow to confirm state-of-the-art results by further experimental
evidences
A Bayesian Analysis of Co-Training Algorithm with Insufficient Views
No full textThe co-training algorithm can be applied if a dataset admits a representation into two different feature sets (two views). However, its optimality is proved only under the conditions a) sufficiency of each view, and b) conditional independence given the class. We address the case where condition a) doesn't hold, as often happens in concrete applications. In such cases the co-training is unable to converge to the optimal Bayesian classifier, because samples added in the training set are not distributed according to the classconditional distributions, even if their assigned label is correct. These results help to better understand the behavior of the co-training algorithm when the classes are only 'statistically' separabl
A study on the performances of dynamic classifier selection based on local accuracy estimation
No full textDynamic classifier selection (DCS) plays a strategic role in the field of multiple classifier systems (MCS). This paper proposes a study on the performances of DCS by Local Accuracy estimation (DCS-LA). To this end, upper bounds against which the performances can be evaluated are proposed. The experimental results on five datasets clearly show the effectiveness of the selection methods based on local accuracy estimates. (c) 2005 Pattern Recognition Societ
A Theoretical and Experimental Analysis of Template Co-update in Biometric Verification Systems
No full textTemplate update in biometric recognition system is aimed to improve the representativeness of available templates in order to make them adaptive to the large intra-class variations characterizing biometrics (e.g. fingerprints and faces). Among others, semi-supervised approaches to template update have been recently proposed. Since the lack of representativeness is due to the impossibility of sampling all possible variations of a given client biometric, these approaches exploit samples submitted during the recognition phase by adding the “highly genuine” ones to the related client gallery. In particular, the template co-update algorithm, which uses the mutual help of two complementary biometric matchers, has shown promising experimental results. However, no theoretical model has been proposed to explain the behaviour of the co-update algorithm and support the experimental results. This is the goal of this paper. Experimental results show the correctness of the proposed theoretical model
- …
