Crash tutorial
On this page, we will apply everything we've covered in the previous tutorials. We will start from scratch and guide you through the final steps of viewing and saving the results.
Video tutorial
Steps
1. Creating the workspace:
When you first open the MedomicsLab application, you see this window where we can view our previously created workspaces. In my case, there is an old workspace, but if you are new to this application, you should click the 'Set workspace' button and choose a folder. It's preferable to choose an empty folder.
2. Uploading Datasets:
For this experience, we will need a dataset to run the pipelines, so we need to add the datasets to the workspace.
For our example, we will use the Diabetes dataset. Make sure to download the files, and once you have them, add them to the DATA folder and refresh the workspace.
MEDfl supports numerical tabular datasets. If you want to test your dataset, ensure it is compatible and clean. It's always preferable to start the tutorial with our provided dataset for consistency.
3. Configuring Database:
When you first open MEDfl, you'll see a popup containing instructions on how to create a configuration file. This file will be used to connect to the MySQL database on your local machine.
To get started, create a file named config.ini
in your workspace and add your connection information, including the username and password.
for more details about configuring the database feel free to check the configuration tutorial page
4. Optimizing hyperparameters:
Before we create the pipeline, we can optimize the model hyperparameters to improve performance.
After the optimization, we can use the results to initialize the central model hyperparameters.
For our example, we will use the Optuna central optimization to optimize the hyperparameters. MEDfl offers various options for optimization, so feel free to check the Optimization tutorial page for more details.
After Having the optimization results we will save them, so we can use them later
5. Creating and running pipelines
for this experience, we will create two configurations and run them simultaneously, each configuration will have a different aggregation algorithm for more details on how to create the pipeline and add the different nodes please check this tutorial page
6. Pipeline results
You can switch between different configuration results and also between different types of results. There are mainly three types of results: Global results, results by node, and compare results.
For more details about the results, refer to the Pipeline Results tutorial page.
7. Merge results
In this section, we will merge two different result files into one file containing both sets of results.
You have the option to add the "Merge Results" node separately from the pipeline, or you can link it to the pipeline so that the pipeline results will be automatically merged with the specified file.
Last updated