Giannis Bekoulis

Giannis Bekoulis

Auto-generated profile, based on publicly available information

Giannis Bekoulis

University of Oslo Vrije Universiteit Brussel -imec University of Copenhagen

Research Focus & Metrics

Academic Career Length

11Years (2016-Present)

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

Representative Publications

View All Publications
1 December 2018 in Expert Systems with Applications
Verified
Journal Article
Closed access
GBGiannis Bekoulis
Giannis Bekoulis
JDJohannes Deleu
Johannes Deleu
TDThomas Demeester
Thomas Demeester
CDChris Develder
Chris Develder

1 January 2018
Verified
Conference Paper
Open access
GBGiannis Bekoulis
Giannis Bekoulis
JDJohannes Deleu
Johannes Deleu
TDThomas Demeester
Thomas Demeester
CDChris Develder
Chris Develder

10 June 2021 in IEEE Transactions on Neural Networks and Learning Systems
Verified
Journal Article
Closed access
LALusine Abrahamyan
Lusine Abrahamyan
YCYiming Chen
Yiming Chen
GBGiannis Bekoulis
Giannis Bekoulis
NDNikos Deligiannis
Nikos Deligiannis
Training deep neural networks on large datasets containing high-dimensional data requires a large amount of computation. A solution to this problem is data-parallel distributed training, where a model is replicated into several computational nodes that have access to different chunks of the data. This approach, however, entails high communication rates and latency because of the computed gradients that need to be shared among nodes at every iteration. The problem becomes more pronounced in the case that there is wireless communication between the nodes (i.e., due to the limited network bandwidth). To address this problem, various compression methods have been proposed, including sparsification, quantization, and entropy encoding of the gradients. Existing methods leverage the intra-node information redundancy, that is, they compress gradients at each node independently. In contrast, we advocate that the gradients across the nodes are correlated and propose methods to leverage this inter-node redundancy to improve compression efficiency. Depending on the node communication protocol (parameter server or ring-allreduce), we propose two instances for the gradient compression that we coin Learned Gradient Compression (LGC). Our methods exploit an autoencoder (i.e., trained during the first stages of the distributed training) to capture the common information that exists in the gradients of the distributed nodes. To constrain the nodes' computational complexity, the autoencoder is realized with a lightweight neural network. We have tested our LGC methods on the image classification and semantic segmentation tasks using different convolutional neural networks (CNNs) [ResNet50, ResNet101, and pyramid scene parsing network (PSPNet)] and multiple datasets (ImageNet, Cifar10, and CamVid). The ResNet101 model trained for image classification on Cifar10 achieved significant compression rate reductions with the accuracy of 93.57%, which is lower than the baseline distributed training with uncompressed gradients only by 0.18%. The rate of the model is reduced by 8095x and 8x compared with the baseline and the state-of-the-art deep gradient compression (DGC) method, respectively.

1 January 2020
Verified
Conference Paper
Open access
PDF file
FNFarhad Nooralahzadeh
Farhad Nooralahzadeh
GBGiannis Bekoulis
Giannis Bekoulis
JBJohannes Bjerva
Johannes Bjerva
IAIsabelle Augenstein
Isabelle Augenstein
Farhad Nooralahzadeh, Giannis Bekoulis, Johannes Bjerva, Isabelle Augenstein. Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP). 2020.

1 July 2018 in Expert Systems with Applications
Verified
Journal Article
Closed access
GBGiannis Bekoulis
Giannis Bekoulis
JDJohannes Deleu
Johannes Deleu
TDThomas Demeester
Thomas Demeester
CDChris Develder
Chris Develder

17 September 2021 in IEEE Access
Verified
Journal Article
Open access
TDTien Huu Do
Tien Huu Do
MBMarc Berneman
Marc Berneman
JPJasabanta Patro
Jasabanta Patro
GBGiannis Bekoulis
Giannis Bekoulis
NDNikos Deligiannis
Nikos Deligiannis
Fake news is a serious problem, which has received considerable attention from both industry and academic communities. Over the past years, many fake news detection approaches have been introduced, and most of the existing methods rely on either news content or the social context of the news dissemination process on social media platforms. In this work, we propose a generic model that is able to take into account both the news content and the social context for the identification of fake news. Specifically, we explore different aspects of the news content by using both shallow and deep representations. The shallow representations are produced with word2vec and doc2vec models while the deep representations are generated via transformer-based models. These representations are able to jointly or separately address four individual tasks, namely bias detection, clickbait detection, sentiment analysis, and toxicity detection. In addition, we make use of graph convolutional neural networks and mean-field layers in order to exploit the underlying structural information of the news articles. That way, we are able to take into account the inherent correlation between the articles by leveraging their social context information. Experiments on widely-used benchmark datasets indicate the effectiveness of the proposed method.