**Question:** What if we want to optimize model configuration or hyper-parameters (e.g. number of image channels, number of hidden layers or learning rate), which might require model training many times. Even if we use the express lane, we need to push and submit each model configuration one at a time, which could be time consuming. Can we do it more efficiently? **Answer**: The purpose of the Express Lane is not to do meaningful training! The description of the Express Lane is available [here](https://www.synapse.org/#!Synapse:syn4224222/wiki/409764). For the rest, you are free to train your model on the submission queue ?Digital Mammography Model Training? as many times as your time quota allows you to do so.