ML-Assisted Resource Allocation Outage Probability: Simple, Closed-Form Approximations.

In this paper, we establish simple and efficient approximations for the outage probability of a single-user multi-resource allocation system that consists of a machine learning (ML) based outage predictor whose task is to assign resources to the user while minimizing outages. We begin by presenting the outage probability expressions for this system. We then propose the approximations to this system’s outage probability using both the sinc function and the zeroth-order Bessel function of the first kind. These approximations are based on naive upper and lower bounds and stem from understanding how the channel samples de-correlate over time. Our results demonstrate that the outage probability indeed lies within the range defined by the bounds. Moreover, the effectiveness of the proposed outage probability approximations are evident as they exhibit a strong alignment with the trend of the outage probability curve. Finally, because of the simplicity of our approximations, they can be calculated in a computationally efficient manner.