Inertial Proximal Deep Learning Alternating Minimization for Efficient Neutral Network Training
This work addresses training efficiency for deep learning practitioners, but it is incremental as it builds on existing alternating minimization methods.
The authors tackled the inefficiency of Stochastic Gradient Descent in deep neural network training by developing an inertial-enhanced Deep Learning Alternating Minimization algorithm with a warm-up technique, resulting in improved training speed as demonstrated on real-world datasets.
In recent years, the Deep Learning Alternating Minimization (DLAM), which is actually the alternating minimization applied to the penalty form of the deep neutral networks training, has been developed as an alternative algorithm to overcome several drawbacks of Stochastic Gradient Descent (SGD) algorithms. This work develops an improved DLAM by the well-known inertial technique, namely iPDLAM, which predicts a point by linearization of current and last iterates. To obtain further training speed, we apply a warm-up technique to the penalty parameter, that is, starting with a small initial one and increasing it in the iterations. Numerical results on real-world datasets are reported to demonstrate the efficiency of our proposed algorithm.