AgEBO-tabular

What is it about?

Deep neural networks are widely used to learn from data. However, developing them requires experience to choose the right neural architecture and training parameters. AgEBO automatically search the configuration of a neural networks and adapts to different high performance computing settings.

Featured Image

Photo by Pietro Jeng on Unsplash

Photo by Pietro Jeng on Unsplash

Why is it important?

This work shows the advantage of tuning both the neural architecture and training hyperparameters at once. It enable the efficient use of many accelerator hardwares (e.g., GPUs). The described methodology is open-sourced in a Python package called DeepHyper highly compatible with the famous Keras/Tensorflow framework.