We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

cs.LG

Change to browse by:

References & Citations

DBLP - CS Bibliography

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Machine Learning

Title: Benchmarking Learnt Radio Localisation under Distribution Shift

Abstract: Deploying radio frequency (RF) localisation systems invariably entails non-trivial effort, particularly for the latest learning-based breeds. There has been little prior work on characterising and comparing how learnt localiser networks can be deployed in the field under real-world RF distribution shifts. In this paper, we present RadioBench: a suite of 8 learnt localiser nets from the state-of-the-art to study and benchmark their real-world deployability, utilising five novel industry-grade datasets. We train 10k models to analyse the inner workings of these learnt localiser nets and uncover their differing behaviours across three performance axes: (i) learning, (ii) proneness to distribution shift, and (iii) localisation. We use insights gained from this analysis to recommend best practices for the deployability of learning-based RF localisation under practical constraints.
Subjects: Machine Learning (cs.LG); Networking and Internet Architecture (cs.NI); Signal Processing (eess.SP)
Cite as: arXiv:2210.01930 [cs.LG]
  (or arXiv:2210.01930v1 [cs.LG] for this version)

Submission history

From: Mo Alloulah [view email]
[v1] Tue, 4 Oct 2022 22:04:45 GMT (5204kb,D)

Link back to: arXiv, form interface, contact.