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

Transformer-based neural network architectures achieve state-of-the-artresults in different domains, from natural language processing (NLP) tocomputer vision (CV). The key idea of Transformers, the attention mechanism,has already led to significant breakthroughs in many areas. The attention hasfound their implementation for time series data as well. However, due to thequadratic complexity of the attention calculation regarding input sequencelength, the application of Transformers is limited by high resource demands.Moreover, their modifications for industrial time series need to be robust tomissing or noised values, which complicates the expansion of the horizon oftheir application. To cope with these issues, we introduce the class ofefficient Transformers named Regularized Transformers (Reguformers). Weimplement the regularization technique inspired by the dropout ideas to improverobustness and reduce computational expenses. The focus in our experiments ison oil&gas data, namely, well logs, a prominent example of multivariate timeseries. The goal is to solve the problems of similarity and representationlearning for them. To evaluate our models for such problems, we work with anindustry-scale open dataset consisting of well logs of more than 20 wells. Theexperiments show that all variations of Reguformers outperform the previouslydeveloped RNNs, classical Transformer model, and robust modifications of itlike Informer and Performer in terms of well-intervals' classification and thequality of the obtained well-intervals' representations. Moreover, thesustainability to missing and incorrect data in our models exceeds that ofothers by a significant margin. The best result that the Reguformer achieves onwell-interval similarity task is the mean PR~AUC score equal to 0.983, which iscomparable to the classical Transformer and outperforms the previous models.

Robust representations of oil wells' intervals via sparse attention mechanism