Feedback

Deep Learning Based Resource Allocation : How Much Training Data is Needed?

We consider artificial neural networks based energy-efficient power control for interference networks. The influence
of different training set sizes and data augmentation is evaluated. It is shown that as few as 15,000 data points obtained from 300 channel realizations are sufficient to adequately predict almost globally optimal power allocations in a 4 user network. Moreover, we observe that, especially for larger scenarios, data augmentation is essential for successful training and far outweighs the effect of increasing the training data set size.

Preview

Category

Conference:
2020 IEEE 21st International Workshop on Signal Processing Advances in Wireless Communications (SPAWC), Atlanta, 26-29 May 2020
Appears in:
IEEE 21st International Workshop on Signal Processing Advances in Wireless Communications (SPAWC)
(2020), p. 1-5
Date Issued:
2020
DOI:
10.24355/dbbs.084-202301111326-0
URN:
urn:nbn:de:gbv:084-2023011113329
Language:
English
Type of Resource:
Text
Extent:
5 Seiten
Place Term:
Piscataway Township, NJ
Publisher:
IEEE
Keywords:
  1. Energy Efficiency
    Typ
    Schlagwort
  2. Resource Allocation
    Typ
    Schlagwort
  3. Neural Networks
    Typ
    Schlagwort
  4. Deep Learning
    Typ
    Schlagwort
  5. Data Augmentation
    Typ
    Schlagwort
DDC:
621.3 Elekto-, magnetische, optische, Kommunikations- und Computertechnik, Elektronik, Beleuchtung
Location:
Publications of TU Braunschweig
Link to Raw Object:
https://doi.org/10.1109/SPAWC48557.2020.9154298
Hosting institution:
05 - Electrical Engineering, Information Technology, Physics

Cite

Citation style:
Could not load citation form.

Access Statistic

Total:
Downloads:
Abtractviews:
Last 12 Month:
Downloads:
Abtractviews:
open graphic

Rights

License Holder: © 2020 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works.

Use and reproduction:
All rights reserved