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

To meet the practical requirements of low latency, low cost, and good privacyin online intelligent services, more and more deep learning models areoffloaded from the cloud to mobile devices. To further deal with cross-devicedata heterogeneity, the offloaded models normally need to be fine-tuned witheach individual user's local samples before being put into real-time inference.In this work, we focus on the fundamental click-through rate (CTR) predictiontask in recommender systems and study how to effectively and efficientlyperform on-device fine-tuning. We first identify the bottleneck issue that eachindividual user's local CTR (i.e., the ratio of positive samples in the localdataset for fine-tuning) tends to deviate from the global CTR (i.e., the ratioof positive samples in all the users' mixed datasets on the cloud for trainingout the initial model). We further demonstrate that such a CTR drift problemmakes on-device fine-tuning even harmful to item ranking. We thus propose anovel label correction method, which requires each user only to change thelabels of the local samples ahead of on-device fine-tuning and can well alignthe locally prior CTR with the global CTR. The offline evaluation results overthree datasets and five CTR prediction models as well as the online A/B testingresults in Mobile Taobao demonstrate the necessity of label correction inon-device fine-tuning and also reveal the improvement over cloud-based learningwithout fine-tuning.

On-Device Model Fine-Tuning with Label Correction in Recommender Systems