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

Humans possess an innate ability to identify and differentiate instances thatthey are not familiar with, by leveraging and adapting the knowledge that theyhave acquired so far. Importantly, they achieve this without deteriorating theperformance on their earlier learning. Inspired by this, we identify andformulate a new, pragmatic problem setting of NCDwF: Novel Class Discoverywithout Forgetting, which tasks a machine learning model to incrementallydiscover novel categories of instances from unlabeled data, while maintainingits performance on the previously seen categories. We propose 1) a method togenerate pseudo-latent representations which act as a proxy for (no longeravailable) labeled data, thereby alleviating forgetting, 2) amutual-information based regularizer which enhances unsupervised discovery ofnovel classes, and 3) a simple Known Class Identifier which aids generalizedinference when the testing data contains instances form both seen and unseencategories. We introduce experimental protocols based on CIFAR-10, CIFAR-100and ImageNet-1000 to measure the trade-off between knowledge retention andnovel class discovery. Our extensive evaluations reveal that existing modelscatastrophically forget previously seen categories while identifying novelcategories, while our method is able to effectively balance between thecompeting objectives. We hope our work will attract further research into thisnewly identified pragmatic problem setting.

Novel Class Discovery without Forgetting