Per-sample prediction intervals for extreme learning machines

Anton Akusok, Yoan Miche, Kaj-Mikael Björk, Amaury Lendasse

Forskningsoutput: TidskriftsbidragArtikelVetenskapligPeer review

Sammanfattning

Prediction intervals in supervised machine learning bound the region where the true outputs of new samples may fall. They are necessary in the task of separating reliable predictions of a trained model from near random guesses, minimizing the rate of false positives, and other problem-specific tasks in applied machine learning. Many real problems have heteroscedastic stochastic outputs, which explains the need of input-dependent prediction intervals. This paper proposes to estimate the input-dependent prediction intervals by a separate extreme learning machine model, using variance of its predictions as a correction term accounting for the model uncertainty. The variance is estimated from the model’s linear output layer with a weighted Jackknife method. The methodology is very fast, robust to heteroscedastic outputs, and handles both extremely large datasets and insufficient amount of training data.
OriginalspråkEngelska
Referentgranskad vetenskaplig tidskriftInternational Journal of Machine Learning and Cybernetics
Sidor (från-till)1-11
Antal sidor11
ISSN1868-8071
DOI
StatusPublicerad - 30.01.2018
MoE-publikationstypA1 Originalartikel i en vetenskaplig tidskrift

Nyckelord

  • 512 Företagsekonomi

Fingeravtryck Fördjupa i forskningsämnen för ”Per-sample prediction intervals for extreme learning machines”. Tillsammans bildar de ett unikt fingeravtryck.

Citera det här