TensorFlow 1.x Deep Learning Cookbook

There's more...

There is another approach called Bayesian optimization, which can also be used to tune hyperparameters. In it, we define an acquisition function along with a Gaussian process. The Gaussian process uses a set of previously evaluated parameters and resulting accuracy to assume about unobserved parameters. The acquisition function using this information suggests the next set of parameters. There is a wrapper available for even gradient-based hyperparameter optimization https://github.com/lucfra/RFHO.