Mitigation of Scaling Trade-offs in Distributed Deep Learning through Multi-Objective Optimization

The potential to solve complex problems along with the performance that deep learning offers has made it gain popularity in the scientific community. Increased performance through scaling creates a challenge related to the trade-off between accuracy and performance. It is mandatory to optimize a set of hyperparameters. In this work, the Multi-Objective Optimization method is presented to find the optimal values of the hyperparameters in a formal way. The expected results is a minimization of the trade-offs.