Creating a New Training Session with a Separate Test Dataset in PrismRCL

Introduction #

PrismRCL allows users to initiate training sessions using not just their training data but also a separate test dataset for evaluating the model’s performance. This method ensures a more robust validation process by testing the model on completely unseen data. This article details the process of starting a new training session with a designated test dataset.

Requirements #

• Dataset Preparation: Your training and test datasets must be correctly prepared and formatted according to PrismRCL specifications.
• PrismRCL Installation: Confirm that PrismRCL is properly installed on your Windows system.

Command Structure #

To begin a training session with a separate test dataset, use the following command:

C:\PrismRCL\PrismRCL.exe chisquared rclticks=15 data=C:\PrismRCL\data\dataset\train-data testdata=C:\PrismRCL\data\dataset\test-data savemodel=C:\PrismRCL\models\model_name.classify log=C:\PrismRCL\logfiles\job_folder stopwhendone

• chisquared: Indicates the evaluation method for this session, optimized for image data analysis.
• rclticks=15: A specific setting for the chisquared evaluation method, applicable only to image data.
• data: The path to your training dataset.
• testdata: The path to your separate test dataset, used for model evaluation.
• savemodel: Specifies where the trained model will be saved.
• log: The directory where logs and result files of the training session will be stored.
• stopwhendone: Commands PrismRCL to automatically shut down after the training session is completed.

Steps for Creating a Training Session with a Test Dataset #

1. Organize Your Data: Ensure your training and test datasets are prepared according to the PrismRCL data structure requirements. This includes organizing images, text, and tabular data properly.
2. Name Your Model: Decide on a unique name for your model. This name will be used in the savemodel path to save your trained model.
3. Specify Datasets: Clearly define the paths to your training and test datasets within the command. Ensure the paths are correct to avoid errors during the training session.
4. Launch the Training Session: Execute the command from the command line, making sure you are in the PrismRCL installation directory or have the executable in your system’s PATH.
5. Monitor Your Training: Keep an eye on the process through the log files generated in the specified log directory. These files will provide insights into the training progress and the final results.

Conclusion #

By employing a separate test dataset for model evaluation, you enhance the credibility of your AI model’s performance metrics. This method provides a clear separation between training and validation phases, ensuring the model’s effectiveness on unseen data. Follow the steps outlined in this guide to successfully execute a training session with a separate test dataset in PrismRCL.