Model order reduction methods applied to neural network training

Abstract Neural networks have emerged as powerful and versatile tools in the field of deep learning. As complexity task increases, so do size architectural network, causing compression techniques to become a focus current research. Parameter truncation can provide significant reduction memory computational complexity. Originating from model order framework, Discrete Empirical Interpolation Method is applied gradient descent training neural analyze for important parameters. The approach various state‐of‐the‐art compared established methods. Further metrics like L 2 Cross‐Entropy Loss, well accuracy rate are reported.