Data Scientist set hyperparameters before training the machine. They are used in order to tune the settings of the Machine Learning model. There are two methods of tuning the hyperparameters and they are Grid Search and Random Search.
In the case of Grid Search, Data Scientists set up a grid of these hyperparameter values and then train a model for each of the combinations. In this method, all the possible combinations of the data are tried and tested, which is not an ideal or efficient method. This method is expensive in terms of both computing power and time.
Random Search, on the other hand, sets up a grid of the hyperparameter values and chooses random combinations in order to train the model. This helps you to control the number of parameters that you need to test. The number of iterations of the search algorithm is based on time and available resources. This method is comparatively less costly and more effective than the Grid Search.
If you wish to learn more about these methods of hyperparameter tuning in Machine Learning, read Machine Learning Tutorial.
Also, you should check out this Machine Learning video tutorial: