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

In this paper, we describe a conceptual design methodology to designdistributed neural network architectures that can perform efficient inferencewithin sensor networks with communication bandwidth constraints. The differentsensor channels are distributed across multiple sensor devices, which have toexchange data over bandwidth-limited communication channels to solve, e.g., aclassification task. Our design methodology starts from a user-definedcentralized neural network and transforms it into a distributed architecture inwhich the channels are distributed over different nodes. The distributednetwork consists of two parallel branches of which the outputs are fused at thefusion center. The first branch collects classification results from local,node-specific classifiers while the second branch compresses each node's signaland then reconstructs the multi-channel time series for classification at thefusion center. We further improve bandwidth gains by dynamically activating thecompression path when the local classifications do not suffice. We validatethis method on a motor execution task in an emulated EEG sensor network andanalyze the resulting bandwidth-accuracy trade-offs. Our experiments show thatthe proposed framework enables up to a factor 20 in bandwidth reduction withminimal loss (up to 2%) in classification accuracy compared to the centralizedbaseline on the demonstrated motor execution task. The proposed method offers away to smoothly transform a centralized architecture to a distributed,bandwidth-efficient network amenable for low-power sensor networks. While theapplication focus of this paper is on wearable brain-computer interfaces, theproposed methodology can be applied in other sensor network-like applicationsas well.

Bandwidth-efficient distributed neural network architectures with application to body sensor networks