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

Despite remarkable advances in automated visual recognition by machines, somevisual tasks remain challenging for machines. Fleuret et al. (2011) introducedthe Synthetic Visual Reasoning Test (SVRT) to highlight this point, whichrequired classification of images consisting of randomly generated shapes basedon hidden abstract rules using only a few examples. Ellis et al. (2015)demonstrated that a program synthesis approach could solve some of the SVRTproblems with unsupervised, few-shot learning, whereas they remainedchallenging for several convolutional neural networks trained with thousands ofexamples. Here we re-considered the human and machine experiments, because theyfollowed different protocols and yielded different statistics. We thus proposeda quantitative reintepretation of the data between the protocols, so that wecould make fair comparison between human and machine performance. We improvedthe program synthesis classifier by correcting the image parsings, and comparedthe results to the performance of other machine agents and human subjects. Wegrouped the SVRT problems into different types by the two aspects of the corecharacteristics for classification: shape specification and location relation.We found that the program synthesis classifier could not solve problemsinvolving shape distances, because it relied on symbolic computation whichscales poorly with input dimension and adding distances into such computationwould increase the dimension combinatorially with the number of shapes in animage. Therefore, although the program synthesis classifier is capable ofabstract reasoning, its performance is highly constrained by the accessibleinformation in image parsings.

Program synthesis performance constrained by non-linear spatial relations in Synthetic Visual Reasoning Test