This project aims at studying how Deep Neural Networks can be used for various forms of tasks in industiral analytics, for example for time-series analysis. This will require explorations of a (hyoer)parameter space where many different predictive neural networks will be trained in parallel and their final performances compared. Since fitting the biases and weights of deep neural networks is computaitonally very demanding, this kind of hyperparameter tuning is quite expensive, it is necessary to perform such experiments using high performance computer centers.