Energy-efficient neural network training through runtime layer freezing, model quantization, and early stopping | Publicación