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

Understanding how the predictions of deep learning models are formed duringthe training process is crucial to improve model performance and fix modeldefects, especially when we need to investigate nontrivial training strategiessuch as active learning, and track the root cause of unexpected trainingresults such as performance degeneration.  In this work, we propose a time-travelling visual solution DeepVisualInsight(DVI), aiming to manifest the spatio-temporal causality while training a deeplearning image classifier. The spatio-temporal causality demonstrates how thegradient-descent algorithm and various training data sampling techniques caninfluence and reshape the layout of learnt input representation and theclassification boundaries in consecutive epochs. Such causality allows us toobserve and analyze the whole learning process in the visible low dimensionalspace. Technically, we propose four spatial and temporal properties and designour visualization solution to satisfy them. These properties preserve the mostimportant information when inverse-)projecting input samples between thevisible low-dimensional and the invisible high-dimensional space, for causalanalyses. Our extensive experiments show that, comparing to baselineapproaches, we achieve the best visualization performance regarding thespatial/temporal properties and visualization efficiency. Moreover, our casestudy shows that our visual solution can well reflect the characteristics ofvarious training scenarios, showing good potential of DVI as a debugging toolfor analyzing deep learning training processes.

DeepVisualInsight: Time-Travelling Visualization for Spatio-Temporal Causality of Deep Classification Training