Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Challenger-Based Combinatorial Bandits for Subcarrier Selection in OFDM Systems

Mohsen Amiri, V Venktesh, Sindri Magnússon

TL;DR

The paper reframes OFDM subcarrier selection as a top-$m$ arm identification problem in stochastic linear bandits and introduces Challenger Champion Sampling (CCS), a shortlist-based, latency-aware approach that focuses on informative champion–challenger comparisons. CCS maintains a small set of current champions and a rotating shortlist of challengers, uses a gap-index to guide sampling, and employs a largest-variance rule to choose measurements, achieving $(\varepsilon,m,\delta)$-PAC guarantees. Theoretical analysis provides a high-probability bound on stopping time via a problem-dependent complexity term, while simulations in a realistic OFDM downlink show CCS yields orders-of-magnitude reductions in gap-index computations and runtime with near-perfect top-$m$ accuracy. The method offers a practical, tunable speed–accuracy trade-off for AI-enabled communication systems and lays groundwork for extensions to RB contiguity, fairness, and hardware-in-the-loop evaluation.

Abstract

This paper investigates the identification of the top-m user-scheduling sets in multi-user MIMO downlink, which is cast as a combinatorial pure-exploration problem in stochastic linear bandits. Because the action space grows exponentially, exhaustive search is infeasible. We therefore adopt a linear utility model to enable efficient exploration and reliable selection of promising user subsets. We introduce a gap-index framework that maintains a shortlist of current estimates of champion arms (top-m sets) and a rotating shortlist of challenger arms that pose the greatest threat to the champions. This design focuses on measurements that yield the most informative gap-index-based comparisons, resulting in significant reductions in runtime and computation compared to state-of-the-art linear bandit methods, with high identification accuracy. The method also exposes a tunable trade-off between speed and accuracy. Simulations on a realistic OFDM downlink show that shortlist-driven pure exploration makes online, measurement-efficient subcarrier selection practical for AI-enabled communication systems.

Challenger-Based Combinatorial Bandits for Subcarrier Selection in OFDM Systems

TL;DR

The paper reframes OFDM subcarrier selection as a top- arm identification problem in stochastic linear bandits and introduces Challenger Champion Sampling (CCS), a shortlist-based, latency-aware approach that focuses on informative champion–challenger comparisons. CCS maintains a small set of current champions and a rotating shortlist of challengers, uses a gap-index to guide sampling, and employs a largest-variance rule to choose measurements, achieving -PAC guarantees. Theoretical analysis provides a high-probability bound on stopping time via a problem-dependent complexity term, while simulations in a realistic OFDM downlink show CCS yields orders-of-magnitude reductions in gap-index computations and runtime with near-perfect top- accuracy. The method offers a practical, tunable speed–accuracy trade-off for AI-enabled communication systems and lays groundwork for extensions to RB contiguity, fairness, and hardware-in-the-loop evaluation.

Abstract

This paper investigates the identification of the top-m user-scheduling sets in multi-user MIMO downlink, which is cast as a combinatorial pure-exploration problem in stochastic linear bandits. Because the action space grows exponentially, exhaustive search is infeasible. We therefore adopt a linear utility model to enable efficient exploration and reliable selection of promising user subsets. We introduce a gap-index framework that maintains a shortlist of current estimates of champion arms (top-m sets) and a rotating shortlist of challenger arms that pose the greatest threat to the champions. This design focuses on measurements that yield the most informative gap-index-based comparisons, resulting in significant reductions in runtime and computation compared to state-of-the-art linear bandit methods, with high identification accuracy. The method also exposes a tunable trade-off between speed and accuracy. Simulations on a realistic OFDM downlink show that shortlist-driven pure exploration makes online, measurement-efficient subcarrier selection practical for AI-enabled communication systems.

Paper Structure

This paper contains 14 sections, 3 theorems, 16 equations, 2 figures, 1 table, 1 algorithm.

Table of Contents

  1. Introduction
  2. System Model and Problem Formulation
  3. Approach: Challenger Champion Sampling
  4. Theoretical Analysis
  5. Simulation Results
  6. Experimental Setup
  7. Baselines
  8. Results
  9. Efficiency
  10. Runtime
  11. Convergence and stability
  12. Number of arm pulls
  13. Discussion
  14. Conclusion

Key Result

Lemma 1

If the SNR estimation error in dB, $\xi_{i,t}$, is sub-Gaussian with proxy variance $\sigma_\xi^2$, then the induced reward noise is sub-Gaussian with proxy variance $\sigma_\eta^2 \;\le\; (\tfrac{\ln 10}{10\ln 2})^{\!2}\sigma_\xi^2$, uniformly over $\gamma_i \ge 0$.

Figures (2)

  • Figure 1: Top-$m$ arm identification runtime comparison between CCS, LinGIFA (LGI), LinUGapE (LUG) and LinGapE (LG).
  • Figure 2: Variation in correctness and latency of CCS with $|C_t|$.

Theorems & Definitions (6)

  • Lemma 1: Sub-Gaussian reward noise under dB-SNR errors
  • proof
  • Definition 1
  • Theorem 1
  • Lemma 2
  • proof