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 Tools annual

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 Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

Recently, uncertainty-aware deep learning methods for multiclass labelingproblems have been developed that provide calibrated class predictionprobabilities and out-of-distribution (OOD) indicators, letting machinelearning (ML) consumers and engineers gauge a model's confidence in itspredictions. However, this extra neural network prediction information ischallenging to scalably convey visually for arbitrary data sources undermultiple uncertainty contexts. To address these challenges, we presentScatterUQ, an interactive system that provides targeted visualizations to allowusers to better understand model performance in context-driven uncertaintysettings. ScatterUQ leverages recent advances in distance-aware neuralnetworks, together with dimensionality reduction techniques, to constructrobust, 2-D scatter plots explaining why a model predicts a test example to be(1) in-distribution and of a particular class, (2) in-distribution but unsureof the class, and (3) out-of-distribution. ML consumers and engineers canvisually compare the salient features of test samples with training examplesthrough the use of a ``hover callback'' to understand model uncertaintyperformance and decide follow up courses of action. We demonstrate theeffectiveness of ScatterUQ to explain model uncertainty for a multiclass imageclassification on a distance-aware neural network trained on Fashion-MNIST andtested on Fashion-MNIST (in distribution) and MNIST digits (out ofdistribution), as well as a deep learning model for a cyber dataset. Wequantitatively evaluate dimensionality reduction techniques to optimize ourcontextually driven UQ visualizations. Our results indicate that the ScatterUQsystem should scale to arbitrary, multiclass datasets. Our code is available athttps://github.com/mit-ll-responsible-ai/equine-webapp

ScatterUQ: Interactive Uncertainty Visualizations for Multiclass Deep Learning Problems