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

Continual learning from data streams is among the most important topics incontemporary machine learning. One of the biggest challenges in this domainlies in creating algorithms that can continuously adapt to arriving data.However, previously learned knowledge may become outdated, as streams evolveover time. This phenomenon is known as concept drift and must be detected tofacilitate efficient adaptation of the learning model. While there exists aplethora of drift detectors, all of them assume that we are dealing withroughly balanced classes. In the case of imbalanced data streams, thosedetectors will be biased towards the majority classes, ignoring changeshappening in the minority ones. Furthermore, class imbalance may evolve overtime and classes may change their roles (majority becoming minority and viceversa). This is especially challenging in the multi-class setting, whererelationships among classes become complex. In this paper, we propose adetailed taxonomy of challenges posed by concept drift in multi-classimbalanced data streams, as well as a novel trainable concept drift detectorbased on Restricted Boltzmann Machine. It is capable of monitoring multipleclasses at once and using reconstruction error to detect changes in each ofthem independently. Our detector utilizes a skew-insensitive loss function thatallows it to handle multiple imbalanced distributions. Due to its trainablenature, it is capable of following changes in a stream and evolving classroles, as well as it can deal with local concept drift occurring in minorityclasses. Extensive experimental study on multi-class drifting data streams,enriched with a detailed analysis of the impact of local drifts and changingimbalance ratios, confirms the high efficacy of our approach.

Concept Drift Detection from Multi-Class Imbalanced Data Streams