Towards automated kernel selection in machine learning systems: A SYCL case study