[D] Hyperparameter optimization best practices

I started my PhD studies recently and currently working on a paper where we train FFNN / CNN / RNN on test execution traces (details are of less importance imo). My question would be, what are the best practices in hyperparameter optimization, and what are the 'must haves' in a scientific paper?

best practices cnn hyperparameter importance machinelearning optimization paper phd practices rnn studies test traces