We have the answers to your questions! - Don't miss our next open house about the data universe!

Hyperparameters: What are they? What are they used for?

- Reading Time: 3 minutes
machine learning engineer training

Machine learning models are powerful tools for solving complex problems, from predicting stock market trends to diagnosing diseases. However, to get the most out of these models, you need to understand the role of hyperparameters and how to optimise them for better performance.

In this article, we will look at the different types of hyperparameters and the techniques used to optimise them. We will also discuss the importance of understanding the relationship between hyperparameters and model performance, and the trade-offs that need to be made.

What is a hyperparameter?

A hyperparameter is a parameter that is used to configure a Machine Learning model. Unlike model parameters, which are learned from training data, hyperparameters must be defined by the user before training the model.

Hyperparameters are generally chosen according to the desired performance of the model, as well as its characteristics and limitations.

Why are hyperparameters important?

Hyperparameters are important because they can have a significant impact on model performance. For example, by using inappropriate values, it is then possible to overfit or underfit the training data.

Overfitting will occur when the model fits the training data too closely and does not generalise well to the test data. This can lead to a drop in the model’s performance on the test data.

Underfitting occurs when the model does not adapt sufficiently to the training data and fails to capture the complex relationships present in the data. This can lead to a drop in the model’s performance on the training and test data.

In summary, it is important to choose the hyperparameters in such a way as to avoid overfitting and underfitting and thus obtain a model that performs well on the test data.

What are the differences between Hyperparameters?

There are several methods for choosing the hyperparameters of a model. Here are a few examples:

Grid search

This parameter consists of trying out all the possible combinations of hyperparameters, training a model for each combination, and choosing the combination that gives the best results.

This method can be very efficient, but it can be very time-consuming if you have a lot of hyperparameters and possible values to try. In addition, it can be expensive in terms of computational resources if you need a lot of model training to find the best combinations.

Despite these drawbacks, grid search remains a popular method for choosing hyperparameters and can be very effective in many situations. If you have the time and computational resources, this is an approach to consider when choosing hyperparameters for your model.

Random search

Random search involves randomly selecting different combinations of hyperparameters and training a model for each combination. This method can be less efficient than grid search, but the advantage is that it is often quicker to implement and can give good results in certain situations.

Bayesian optimisation

Consists of using a probability distribution on the hyperparameters and updating this distribution according to the results obtained when training the model. This method can be more efficient than grid search and random search, but it generally requires the use of specialised tools and can be more complex to implement.

Experimentation and trial and error

This involves iteratively trying out different hyperparameter values and adjusting the hyperparameters according to the results obtained. This method can be useful when you have a good understanding of the hyperparameters and their impact on the model, but it can be less effective than the other methods when you don’t know which hyperparameters are the most important.

Ultimately, the choice of hyperparameters depends on many factors, such as the complexity of the data, the desired performance and the resources available. It is advisable to test several approaches to find the hyperparameters that give the best results for your model.


Hyperparameters are parameters that are used to configure a Machine Learning model and that can have a significant impact on its performance. It is important to choose hyperparameters carefully to avoid overfitting and underfitting and to obtain a high-performance model.

There are several methods for choosing hyperparameters, each with its own advantages and disadvantages. Ultimately, the choice of hyperparameters depends on many factors and may require some trial and error to find the best values.

You are not available?

Leave us your e-mail, so that we can send you your new articles when they are published!
icon newsletter


Get monthly insider insights from experts directly in your mailbox