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

Our goal is to improve reliability of Machine Learning (ML) systems deployedin the wild. ML models perform exceedingly well when test examples are similarto train examples. However, real-world applications are required to perform onany distribution of test examples. Current ML systems can fail silently on testexamples with distribution shifts. In order to improve reliability of ML modelsdue to covariate or domain shift, we propose algorithms that enable models to:(a) generalize to a larger family of test distributions, (b) evaluate accuracyunder distribution shifts, (c) adapt to a target distribution. We study causesof impaired robustness to domain shifts and present algorithms for trainingdomain robust models. A key source of model brittleness is due to domainoverfitting, which our new training algorithms suppress and instead encouragedomain-general hypotheses. While we improve robustness over standard trainingmethods for certain problem settings, performance of ML systems can still varydrastically with domain shifts. It is crucial for developers and stakeholdersto understand model vulnerabilities and operational ranges of input, whichcould be assessed on the fly during the deployment, albeit at a great cost.Instead, we advocate for proactively estimating accuracy surfaces over anycombination of prespecified and interpretable domain shifts for performanceforecasting. We present a label-efficient estimation to address estimation overa combinatorial space of domain shifts. Further, when a model's performance ona target domain is found to be poor, traditional approaches adapt the modelusing the target domain's resources. Standard adaptation methods assume accessto sufficient labeled resources, which may be impractical for deployed models.We initiate a study of lightweight adaptation techniques with only unlabeleddata resources with a focus on language applications.

Robustness, Evaluation and Adaptation of Machine Learning Models in the Wild