Zakaria Elberrichi

Zakaria Elberrichi

Auto-generated profile, based on publicly available information

Zakaria Elberrichi

EEDIS Laboraory, Djillali Liabes University, Sidi Bel Abbes, Algeria

Research Focus & Metrics

Academic Career Length

16Years (2011-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
1 July 2020 in International Journal of Swarm Intelligence Research
Verified
Journal Article
Closed access
PDF file
NDNassima Dif
Nassima Dif
ZEZakaria Elberrichi
Zakaria Elberrichi
Deep learning is one of the most commonly used techniques in computer-aided diagnosis systems. Their exploitation for histopathological image analysis is important because of the complex morphology of whole slide images. However, the main limitation of these methods is the restricted number of available medical images, which can lead to an overfitting problem. Many studies have suggested the use of static ensemble learning methods to address this issue. This article aims to propose a new dynamic ensemble deep learning method. First, it generates a set of models based on the transfer learning strategy from deep neural networks. Then, the relevant subset of models is selected by the particle swarm optimization algorithm and combined by voting or averaging methods. The proposed approach was tested on a histopathological dataset for colorectal cancer classification, based on seven types of CNNs. The method has achieved accurate results (94.52%) by the Resnet121 model and the voting strategy, which provides important insights into the efficiency of dynamic ensembling in deep learning.

1 March 2012
Verified
Conference Paper
Closed access
YEYoussef Elmir
Youssef Elmir
ZEZakaria Elberrichi
Zakaria Elberrichi
RAReda Adjoudj
Reda Adjoudj
Feature level based monomodal biometric systems perform person recognition based on a multiple sources of biometric information and are affected by problems like integration of evidence obtained from multiple cues and normalization of features codes since they are heterogeneous, in addition of monomodal biometric systems problems like noisy sensor data, non-universality and lack of individuality of the chosen biometric trait, absence of an invariant representation for the biometric trait and susceptibility to circumvention. Some of these problems can be alleviated by using multimodal biometric systems that consolidate evidence from scores of multiple biometric systems. In this work, we address two important issues related to score level fusion. We have studied the performance of a score level fusion based multimodal biometric system against different monomodal biometric system based on voice, fingerprint modalities and a bimodal biometric system based on feature level fusion of the same modalities. These systems have been evaluated in terms of their efficiency and identification rate on a close group from the test data. These results are shown using cumulative match characteristic curve.


1 April 2018 in International Journal of Strategic Information Technology and Applications
Verified
Journal Article
Closed access
PDF file
NDNassima Dif
Nassima Dif
ZEZakaria Elberrichi
Zakaria Elberrichi
Instance selection and feature selection are important steps in the data mining process. They help reduce the excessive number of instances and features. The purpose of this reduction is to eliminate the noisy and redundant instances and features in order to improve the classifiers performance. Various related works in the literature proves that metaheuristics can resolve the problem of instance and feature selection. In this article, the authors propose a new instance selection approach based on a Multi- Verse Optimizer algorithm (MVOIS), to reduce the run time and improve the performance of the one nearest neighbor classifier (1NN). This article tested the proposed approach on 31 datasets from the UCI repository and performed three more pre-process ISFS, FS and FSIS. The comparative study illustrates the efficiency of ISFS and FSIS compared to FS and IS. ISFS achieved 100% accuracy for labor and iris datasets.

1 October 2020 in International Journal of Cognitive Informatics and Natural Intelligence
Verified
Journal Article
Open access
PDF file
NDNassima Dif
Nassima Dif
ZEZakaria Elberrichi
Zakaria Elberrichi
Deep learning methods are characterized by their capacity to learn data representation compared to the traditional machine learning algorithms. However, these methods are prone to overfitting on small volumes of data. The objective of this research is to overcome this limitation by improving the generalization in the proposed deep learning framework based on various techniques: data augmentation, small models, optimizer selection, and ensemble learning. For ensembling, the authors used selected models from different checkpoints and both voting and unweighted average methods for combination. The experimental study on the lymphomas histopathological dataset highlights the efficiency of the MobileNet2 network combined with the stochastic gradient descent (SGD) optimizer in terms of generalization. The best results have been achieved by the combination of the best three checkpoint models (98.67% of accuracy). These findings provide important insights into the efficiency of the checkpoint ensemble learning method for histopathological image classification.

1 April 2020 in International Journal of Service Science, Management, Engineering, and Technology
Verified
Journal Article
Closed access
PDF file
NDNassima Dif
Nassima Dif
ZEZakaria Elberrichi
Zakaria Elberrichi
This article presents a new fine-tuning framework for histopathological images analysis. Despite the most common solutions where the ImageNet models are reused for image classification, this research sets out to perform an intra-domain fine tuning between the trained models on the histopathological images. The purpose is to take advantage of the hypothesis on the efficiency of transfer learning between non-distant datasets and to examine for the first time these suggestions on the histopathological images. The Inception-v3 convolutional neural network architecture, six histopathological source datasets, and four target sets as base modules were used in this article. The obtained results reveal the importance of the pre-trained histopathological models compared to the ImageNet model. In particular, the ICIAR 2018-A presented a high-quality source model for the various target tasks due to its capacity in generalization. Finally, the comparative study with the other literature results shows that the proposed method achieved the best results on both CRC (95.28%) and KIMIA-PATH (98.18%) datasets.