Online Learning of Any-to-Any Path Loss Maps

Gutierrez-Estevez, M. A.; Kasparick, Martin; Stańczak, Sławomir

doi:10.1109/lcomm.2021.3050252

Cited by 3 publications

(6 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…different traffic profiles or change in the underlying map. In the seeding publication of this work [19], we overcome these limitations by proposing an online algorithm which, upon arrival of new measurements, obtains new estimates of both the SLF and the model. To do this, the online algorithm implements a "descent" version of the generalized alternating minimization (gAM) [20], i.e.…”

Section: A Prior Artmentioning

confidence: 99%

“…This strategy results in a nonconvex optimization problem, but it is marginally convex, i.e., the problem becomes convex if a subset of variables is fixed. In contrast to [8], [9], [19], we constrain the updates to remain close to the original model.…”

Section: B Contributionsmentioning

confidence: 99%

Section: B Contributionsmentioning

confidence: 99%

“…2) The problem of learning the model derived from the TPT would be extremely ill-posed, so we propose a non-linear kernel approach based on the radial basis function (RBF) similar to [8], [9], [19]. 3) As in [5], [9], [19], the structure representing the SLF is assumed to be group-sparse, so we consider the least squares problem regularized by the elastic net [21] for the reconstruction of the SLF. 4) Similar to the seeding paper of this work [19], we define a majorizing function that upperbounds the original objective function, i.e., the new function is grater or equal than the original one in its entire domain.…”

Section: B Contributionsmentioning

confidence: 99%

“…3) As in [5], [9], [19], the structure representing the SLF is assumed to be group-sparse, so we consider the least squares problem regularized by the elastic net [21] for the reconstruction of the SLF. 4) Similar to the seeding paper of this work [19], we define a majorizing function that upperbounds the original objective function, i.e., the new function is grater or equal than the original one in its entire domain. This strategy is known from stochastic approximation [22] and has been exploited for online learning in different application domains [5], [23], [24].…”

Section: B Contributionsmentioning

confidence: 99%

See 4 more Smart Citations

Hybrid Model and Data Driven Algorithm for Online Learning of Any-to-Any Path Loss Maps

Gutierrez-Estevez,

Kasparick,

Cavalvante

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

Learning any-to-any (A2A) path loss maps, where the objective is the reconstruction of path loss between any two given points in a map, might be a key enabler for many applications that rely on device-to-device (D2D) communication. Such applications include machine-type communications (MTC) or vehicle-to-vehicle (V2V) communications. Current approaches for learning A2A maps are either model-based methods, or pure data-driven methods. Model-based methods have the advantage that they can generate reliable estimations with low computational complexity, but they cannot exploit information coming from data. Pure data-driven methods can achieve good performance without assuming any physical model, but their complexity and their lack of robustness is not acceptable for many applications. In this paper, we propose a novel hybrid model and data-driven approach that fuses information obtained from datasets and models in an online fashion. To that end, we leverage the framework of stochastic learning to deal with the sequential arrival of samples and propose an online algorithm that alternatively and sequentially minimizes the original nonconvex problem. A proof of convergence is presented, along with experiments based firstly on synthetic data, and secondly on a more realistic dataset for V2X, with both experiments showing promising results.

show abstract

Section: A Prior Artmentioning

confidence: 99%