Utilising convolutions of random functions to realise function calculation via a physical channel

Sigg, Stephan; Jakimovski, Predrag; Ji, Yusheng; Beigl, Michael

doi:10.1109/spawc.2013.6612067

Cited by 2 publications

(2 citation statements)

References 20 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We propose probabilistic collaboration to achieve physical-layer aggregation for non-synchronized distributed devices [106]. In particular, we utilize Poissondistributed burst sequences to compute the sum of the weighted feature inputs n i=1 w i x i during simultaneous transmission.…”

Section: Low Overhead Device Collaborationmentioning

confidence: 99%

“…The coordinator decodes a received burst sequence by counting the number of bursts in a pre-defined interval of length t and thereby estimating the mean M = n i=1 w i x i of the distribution encoded in the superimposed signal sequence [106]. From this non-reversibly aggregated weighted sum of the individual features, the coordinator then computes the model inference as h(x) = 1 1+e w T x + c. While burst collisions can not be avoided in the scheme, their probability, and hence the accuracy in the estimation of the distribution at the receiver can be controlled repeating the estimation multiple times (for T >> t) and by proper choice of k, t and T [106].…”

Section: Model Inference Via Probabilistic Collaborationmentioning

confidence: 99%

See 1 more Smart Citation

Camouflage Learning: Feature Value Obscuring Ambient Intelligence for Constrained Devices

Nguyen

Sigg

Lietzén

et al. 2023

IEEE Trans. on Mobile Comput.

View full text Add to dashboard Cite

Ambient intelligence demands collaboration schemes for distributed constrained devices which are not only highly energy efficient with respect to distributed sensing, processing and communication, but which also respect data privacy. Traditional algorithms for distributed processing suffer in Ambient intelligence domains either from limited data privacy, or from their excessive processing demands for constrained distributed devices. In this paper, we present Camouflage learning, a distributed machine learning scheme that obscures the trained model via probabilistic collaboration using physical-layer computation offloading and demonstrate the feasibility of the approach on backscatter communication prototypes and in comparison with federated learning, a popular distributed learning scheme. We show that Camouflage learning is more energy efficient than traditional schemes and that it requires less communication overhead while reducing the computation load through physical-layer computation offloading. The scheme is synchronization-agnostic and thus appropriate for sharply constrained, synchronization-incapable devices. We demonstrate model training and inference on four distinct datasets and investigate the performance of the scheme with respect to communication range, impact of challenging communication environments, power consumption, and the backscatter hardware prototype.

show abstract

Section: Low Overhead Device Collaborationmentioning

confidence: 99%

Section: Model Inference Via Probabilistic Collaborationmentioning

confidence: 99%