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

Both biological and artificial neural networks inherently balance theirperformance with their operational cost, which balances their computationalabilities. Typically, an efficient neuromorphic neural network is one thatlearns representations that reduce the redundancies and dimensionality of itsinput. This is for instance achieved in sparse coding, and sparserepresentations derived from natural images yield representations that areheterogeneous, both in their sampling of input features and in the variance ofthose features. Here, we investigated the connection between natural images'structure, particularly oriented features, and their corresponding sparsecodes. We showed that representations of input features scattered acrossmultiple levels of variance substantially improve the sparseness and resilienceof sparse codes, at the cost of reconstruction performance. This echoes thestructure of the model's input, allowing to account for the heterogeneouslyaleatoric structures of natural images. We demonstrate that learning kernelfrom natural images produces heterogeneity by balancing between approximate anddense representations, which improves all reconstruction metrics. Using aparametrized control of the kernels' heterogeneity used by a convolutionalsparse coding algorithm, we show that heterogeneity emphasizes sparseness,while homogeneity improves representation granularity. In a broader context,these encoding strategy can serve as inputs to deep convolutional neuralnetworks. We prove that such variance-encoded sparse image datasets enhancecomputational efficiency, emphasizing the benefits of kernel heterogeneity toleverage naturalistic and variant input structures and possible applications toimprove the throughput of neuromorphic hardware.

Kernel Heterogeneity Improves Sparseness of Natural Images Representations