Brief description of the project, its purpose, and its features.
Before you begin, ensure you have met the following requirements:
- Programming Language: MATLAB
Follow these steps to run the project:
-
Prepare the Tabular Dataset
Tabular datasets in the appropriate format (e.g. csv) should be located in the
TabularDatasetsdirectory. You can assign the path of a tabular dataset that you want to use to the variable, as shown in the example below.tabularFile = 'TabularDatasets\RiceMSCDataset.csv';
-
Create an Instance of the NuDIT
Initialize the NuDIT object with your dataset file.
nudit = NuDIT(tabularFile);
During the transformation phase of the tabular dataset, you will see a progress bar as shown below. The duration of this process is directly proportional to the number of records in the tabular dataset. Once the transformation process is complete, the next stage is automatically started.
Important: If you have done the transformation process once for a tabular dataset, comment this line of code in subsequent runs. Otherwise, the transformation will be done every time.
-
Transform numerical data into images suitable for deep learning models.
nudit.numToImgTransform();
Set Parameters
-
Define the image size and number of folds for cross-validation. For example, set the image size to 32 pixels and use 5 folds.
imageSize = 32; kfold = 5;
-
Resize images and split the data into k folds for cross-validation.
Prepare your image dataset for k-fold cross-validation
nudit.prepareData(imageSize, kfold);
Train the Network
The default training options for DAG-Net are as follows:
- Optimizer: Stochastic Gradient Descent with Momentum (SGDM)
- Mini-Batch Size: 16
- Maximum Epochs:
epochs(variable) - Execution Environment: CPU
Here is an example code snippet showing how the default settings are configured:
trainingOptions('sgdm', ...
'MiniBatchSize', 16, 'MaxEpochs', epochs, 'ExecutionEnvironment', 'cpu');These default settings define the core parameters used during the training process of the model. If you wish to customize these training options, you can modify the trainingOptions function within the runNetwork method of the NuDIT class.
-
Execute the DAG-Net network with a specified number of epochs.
epochs = 10; result = nudit.runNetwork(epochs);
The results (image dataset) will be saved in the Exports directory specified in your script. Check this directory for the results and any log files.
If you encounter issues, verify the following:
- The dataset file exists in the
TabularDatasetsdirectory. - All necessary packages and toolboxes are installed and properly configured.
For contributions, please follow the guidelines in the CONTRIBUTING.md file.
This project is licensed under the MIT License. See the LICENSE file for more details.
For further assistance, please contact:
- Email: aelen@bandirma.edu.tr
- GitHub: abdullahelen