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

As object recognition becomes an increasingly common ML task, and recentresearch demonstrating CNNs vulnerability to attacks and small imageperturbations necessitate fully understanding the foundations of objectrecognition. We focus on understanding the mechanisms behind how neuralnetworks generalize to spatial transformations of complex objects. While humansexcel at discriminating between objects shown at new positions, orientations,and scales, past results demonstrate that this may be limited to familiarobjects - humans demonstrate low tolerance of spatial-variances forpurposefully constructed novel objects. Because training artificial neuralnetworks from scratch is similar to showing novel objects to humans, we seek tounderstand the factors influencing the tolerance of CNNs to spatialtransformations. We conduct a thorough empirical examination of sevenConvolutional Neural Network (CNN) architectures. By training on a controlledface image dataset, we measure model accuracy across different degrees of 5transformations: position, size, rotation, Gaussian blur, and resolutiontransformation due to resampling. We also examine how learning strategy affectsgeneralizability by examining how different amounts of pre-training have onmodel robustness. Overall, we find that the most significant contributor totransformation invariance is pre-training on a large, diverse image dataset.Moreover, while AlexNet tends to be the least robust network, VGG and ResNetarchitectures demonstrate higher robustness for different transformations.Along with kernel visualizations and qualitative analyses, we examinedifferences between learning strategy and inherent architectural properties incontributing to invariance of transformations, providing valuable informationtowards understanding how to achieve greater robustness to transformations inCNNs.

The Effect of Learning Strategy versus Inherent Architecture Properties  on the Ability of Convolutional Neural Networks to Develop Transformation  Invariance