Data-Driven Symbol Detection Via Model-Based Machine Learning

We present a data-driven framework to symbol detection design that combines machine learning (ML) and model-based algorithms. The resulting data-driven receivers are most suitable for systems where the underlying channel models are poorly understood, highly complex, or do not well-capture the underlying physics. Our approach is unique in that it only replaces the channel-model-based computations with dedicated neural networks that can be trained from a small amount of data, while keeping the general algorithm intact. Our results demonstrate that these techniques can yield performance close to that of model-based algorithms with perfect model knowledge without knowing the exact channel model or state.