Training a model simply means learning (determining) good values for all the weights and the bias from labeled examples.

Training a model simply means learning (determining) good values for all the weights and the bias from labeled examples. In supervised learning, a machine learning algorithm builds a model by examining many examples and attempting to find a model that minimizes loss; this process is called empirical risk minimization.

On Dataleon platform you can train a model in the Knowledge Vision part. On the whole the platform consists of four parts. The other three parts of the platform are Pre-Trained Models, Datalabs and Analytics.

In Pre-Trained Models part of the platform you will find our ready-to-use models for data detection, extracton, analysis and classification.In Datalabs section you can create data in order to train your own model. In Knowledge Vision part of the platform you can train your model with your data (either personal or generated in Datalabs).And finally in Analytics section you can analyze your inference request on production.

In order to train a model you will need to upload some data. These data will be further used for training the model. You can easily do it in Datalabs/Presets.

Afterwards you can create a project in Datalabs/Projects with these data.

When creating a project in Datalabs, you can label the data previously uploaded in Presets subpart. Below you can see a screenshot of Labelling Editor, i.e a labelling interface used to label your datasets.

In case you do not have any data for labelling , you can generate them in Scenes Editor. These data can be also used later for labelling.

When the project is created you can view all the information about it in Settings.

After working with data, you can finally create and train a model. To do this you will be asked to follow 6 steps in Knowledge Vision section.

On step 4 you will be asked to select either a basic or an expert mode. If your are not a data scientist, select a basic mode. If you select an expert mode, you will need to specify your model and its training parameters, hyperparameters, data augmention and evaluation metrics.

In the process of training you can view all the detailed information, such as logs, checkpoints and metrics.

After that you click Upload to upload your data and you can test the model with your real data.

Training a model can be easy. Just try!!!

So, why to choose Dataleon platform for training a model?

• No MLOps or DevOps are needed for deployment.
• You can select Expert mode if you want to.
• Dataleon is no-code AI platform, which designs a custom solution to meet your demands.

‍

‍