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

The integration of machine learning models in various real-world applicationsis becoming more prevalent to assist humans in their daily decision-makingtasks as a result of recent advancements in this field. However, it has beendiscovered that there is a tradeoff between the accuracy and fairness of thesedecision-making tasks. In some cases, these AI systems can be unfair byexhibiting bias or discrimination against certain social groups, which can havesevere consequences in real life. Inspired by one of the most well-known humanlearning skills called grouping, we address this issue by proposing a novelmachine learning framework where the ML model learns to group a diverse set ofproblems into distinct subgroups to solve each subgroup using its specificsub-model. Our proposed framework involves three stages of learning, which areformulated as a three-level optimization problem: (i) learning to groupproblems into different subgroups; (ii) learning group-specific sub-models forproblem-solving; and (iii) updating group assignments of training examples byminimizing the validation loss. These three learning stages are performedend-to-end in a joint manner using gradient descent. To improve fairness andaccuracy, we develop an efficient optimization algorithm to solve thisthree-level optimization problem. To further reduce the risk of overfitting insmall datasets, we incorporate domain adaptation techniques in the second stageof training. We further apply our method to neural architecture search.Extensive experiments on various datasets demonstrate our method'seffectiveness and performance improvements in both fairness and accuracy. Ourproposed Learning by Grouping can reduce overfitting and achievestate-of-the-art performances with fixed human-designed network architecturesand searchable network architectures on various datasets.

Learning by Grouping: A Multilevel Optimization Framework for Improving Fairness in Classification without Losing Accuracy