Differentiable High-Order Markov Models for Spectrum Prediction

TL;DR

This paper revisits high-order Markov models for spectrum prediction, introducing a gradient-based fine-tuning framework that enhances classical probabilistic models with modern learning techniques, showing competitive results with deep learning.

Contribution

It presents a novel framework for high-order Markov models that addresses model mismatch and enables gradient-based fine-tuning, bridging classical and modern approaches.

Findings

High-order Markov models perform competitively with deep learning methods.

The proposed framework effectively handles dataset constraints and outliers.

Simulations on real Wi-Fi traffic validate the approach's robustness.

Abstract

The advent of deep learning and recurrent neural networks revolutionized the field of time-series processing. Therefore, recent research on spectrum prediction has focused on the use of these tools. However, spectrum prediction, which involves forecasting wireless spectrum availability, is an older field where many "classical" tools were considered around the 2010s, such as Markov models. This work revisits high-order Markov models for spectrum prediction in dynamic wireless environments. We introduce a framework to address mismatches between sensing length and model order as well as state-space complexity arising with large order. Furthermore, we extend this Markov framework by enabling fine-tuning of the probability transition matrix through gradient-based supervised learning, offering a hybrid approach that bridges probabilistic modeling and modern machine learning. Simulations on real-world Wi-Fi traffic demonstrate the competitive performance of high-order Markov models compared to deep learning methods, particularly in scenarios with constrained datasets containing outliers.