English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

In this study, we present a dynamic graph representation learning model onweighted graphs to accurately predict the network capacity of connectionsbetween viewers in a live video streaming event. We propose EGAD, a neuralnetwork architecture to capture the graph evolution by introducing aself-attention mechanism on the weights between consecutive graph convolutionalnetworks. In addition, we account for the fact that neural architecturesrequire a huge amount of parameters to train, thus increasing the onlineinference latency and negatively influencing the user experience in a livevideo streaming event. To address the problem of the high online inference of avast number of parameters, we propose a knowledge distillation strategy. Inparticular, we design a distillation loss function, aiming to first pretrain ateacher model on offline data, and then transfer the knowledge from the teacherto a smaller student model with less parameters. We evaluate our proposed modelon the link prediction task on three real-world datasets, generated by livevideo streaming events. The events lasted 80 minutes and each viewer exploitedthe distribution solution provided by the company Hive Streaming AB. Theexperiments demonstrate the effectiveness of the proposed model in terms oflink prediction accuracy and number of required parameters, when evaluatedagainst state-of-the-art approaches. In addition, we study the distillationperformance of the proposed model in terms of compression ratio for differentdistillation strategies, where we show that the proposed model can achieve acompression ratio up to 15:100, preserving high link prediction accuracy. Forreproduction purposes, our evaluation datasets and implementation are publiclyavailable at https://stefanosantaris.github.io/EGAD.

EGAD: Evolving Graph Representation Learning with Self-Attention and Knowledge Distillation for Live Video Streaming Events