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

Distributed machine learning training is one of the most common and importantworkloads running on data centers today, but it is rarely executed alone.Instead, to reduce costs, computing resources are consolidated and shared bydifferent applications. In this scenario, elasticity and proper load balancingare vital to maximize efficiency, fairness, and utilization. Currently, mostdistributed training frameworks do not support the aforementioned properties. Afew exceptions that do support elasticity, imitate generic distributedframeworks and use micro-tasks. In this paper we illustrate that micro-tasksare problematic for machine learning applications, because they require a highdegree of parallelism which hinders the convergence of distributed training ata pure algorithmic level (i.e., ignoring overheads and scalabilitylimitations). To address this, we propose Chicle, a new elastic distributedtraining framework which exploits the nature of machine learning algorithms toimplement elasticity and load balancing without micro-tasks. We use Chicle totrain deep neural network as well as generalized linear models, and show thatChicle achieves performance competitive with state of the art rigid frameworks,while efficiently enabling elastic execution and dynamic load balancing.

Addressing Algorithmic Bottlenecks in Elastic Machine Learning with Chicle