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 randomized or cross-validated split of training and testing sets has beenadopted as the gold standard of machine learning for decades. The establishmentof these split protocols are based on two assumptions: (i)-fixing the datasetto be eternally static so we could evaluate different machine learningalgorithms or models; (ii)-there is a complete set of annotated data availableto researchers or industrial practitioners. However, in this article, we intendto take a closer and critical look at the split protocol itself and point outits weakness and limitation, especially for industrial applications. In manyreal-world problems, we must acknowledge that there are numerous situationswhere assumption (ii) does not hold. For instance, for interdisciplinaryapplications like drug discovery, it often requires real lab experiments toannotate data which poses huge costs in both time and financial considerations.In other words, it can be very difficult or even impossible to satisfyassumption (ii). In this article, we intend to access this problem andreiterate the paradigm of active learning, and investigate its potential onsolving problems under unconventional train/test split protocols. We furtherpropose a new adaptive active learning architecture (AAL) which involves anadaptation policy, in comparison with the traditional active learning that onlyunidirectionally adds data points to the training pool. We primarily justifyour points by extensively investigating an interdisciplinary drug-proteinbinding problem. We additionally evaluate AAL on more conventional machinelearning benchmarking datasets like CIFAR-10 to demonstrate thegeneralizability and efficacy of the new framework.

A critical look at the current train/test split in machine learning