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 Tools annual

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 Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

Pre-training on graph neural networks (GNNs) aims to learn transferableknowledge for downstream tasks with unlabeled data, and it has recently becomean active research area. The success of graph pre-training models is oftenattributed to the massive amount of input data. In this paper, however, weidentify the curse of big data phenomenon in graph pre-training: more trainingdata do not necessarily lead to better downstream performance. Motivated bythis observation, we propose a better-with-less framework for graphpre-training: fewer, but carefully chosen data are fed into a GNN model toenhance pre-training. The proposed pre-training pipeline is called thedata-active graph pre-training (APT) framework, and is composed of a graphselector and a pre-training model. The graph selector chooses the mostrepresentative and instructive data points based on the inherent properties ofgraphs as well as predictive uncertainty. The proposed predictive uncertainty,as feedback from the pre-training model, measures the confidence level of themodel in the data. When fed with the chosen data, on the other hand, thepre-training model grasps an initial understanding of the new, unseen data, andat the same time attempts to remember the knowledge learned from previous data.Therefore, the integration and interaction between these two components form aunified framework (APT), in which graph pre-training is performed in aprogressive and iterative way. Experiment results show that the proposed APT isable to obtain an efficient pre-training model with fewer training data andbetter downstream performance.

Better with Less: A Data-Active Perspective on Pre-Training Graph Neural Networks