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

Recent advancements in deep learning have been primarily driven by the use oflarge models trained on increasingly vast datasets. While neural scaling lawshave emerged to predict network performance given a specific level ofcomputational resources, the growing demand for expansive datasets raisesconcerns. To address this, a new research direction has emerged, focusing onthe creation of synthetic data as a substitute. In this study, we investigatehow neural networks exhibit shape bias during training on synthetic datasets,serving as an indicator of the synthetic data quality. Specifically, ourfindings indicate three key points: (1) Shape bias varies across networkarchitectures and types of supervision, casting doubt on its reliability as apredictor for generalization and its ability to explain differences in modelrecognition compared to human capabilities. (2) Relying solely on shape bias toestimate generalization is unreliable, as it is entangled with diversity andnaturalism. (3) We propose a novel interpretation of shape bias as a tool forestimating the diversity of samples within a dataset. Our research aims toclarify the implications of using synthetic data and its associated shape biasin deep learning, addressing concerns regarding generalization and datasetquality.

Harnessing Synthetic Datasets: The Role of Shape Bias in Deep Neural Network Generalization