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

Web-scale ranking systems at Meta serving billions of users is complex.Improving ranking models is essential but engineering heavy. Automated MachineLearning (AutoML) can release engineers from labor intensive work of tuningranking models; however, it is unknown if AutoML is efficient enough to meettight production timeline in real-world and, at the same time, bring additionalimprovements to the strong baselines. Moreover, to achieve higher rankingperformance, there is an ever-increasing demand to scale up ranking models toeven larger capacity, which imposes more challenges on the efficiency. Thelarge scale of models and tight production schedule requires AutoML tooutperform human baselines by only using a small number of model evaluationtrials (around 100). We presents a sampling-based AutoML method, focusing onneural architecture search and hyperparameter optimization, addressing thesechallenges in Meta-scale production when building large capacity models. Ourapproach efficiently handles large-scale data demands. It leverages alightweight predictor-based searcher and reinforcement learning to explore vastsearch spaces, significantly reducing the number of model evaluations. Throughexperiments in large capacity modeling for CTR and CVR applications, we showthat our method achieves outstanding Return on Investment (ROI) versus humantuned baselines, with up to 0.09% Normalized Entropy (NE) loss reduction or$25\%$ Query per Second (QPS) increase by only sampling one hundred models onaverage from a curated search space. The proposed AutoML method has alreadymade real-world impact where a discovered Instagram CTR model with up to -0.36%NE gain (over existing production baseline) was selected for large-scale onlineA/B test and show statistically significant gain. These production resultsproved AutoML efficacy and accelerated its adoption in ranking systems at Meta.

AutoML for Large Capacity Modeling of Meta's Ranking Systems