Abstract Wave–particle interactions are a fundamental driver of electron radiation belt dynamics. Quantifying their effects through quasi‐linear theory requires diffusion coefficients, but their direct evaluation involves nested integrations and is computationally expensive, limiting their use in real‐time applications. Here we develop a fully connected neural network to efficiently estimate diffusion coefficients using a data‐driven approach. The model produces coefficients more than an order of magnitude faster than conventional numerical methods on a single CPU core, and achieves significantly larger speedup when run on a GPU commonly used in modern machine‐learning workflows. To validate its reliability, we use the model‐predicted coefficients in a quasi‐linear simulation, which produces results in close agreement with those obtained using precisely calculated values. These results demonstrate that data‐driven approaches provide a practical path for incorporating diffusion coefficients into real‐time radiation belt forecasting, and more broadly illustrate how such approaches can accelerate computationally intensive physics‐based models.
Tan et al. (Fri,) studied this question.