When training and testing CTGAN, as well as other deep learning models, we need to ensure that the output of each layer is as expected, as well as adding integration tests to validate the overall performance and stability of the model.

Objectives Record the output of each layer: Capture the intermediate output of each layer during the forward process of the model and analyse it. Verify the plausibility of the output: Check the distribution, shape, mean, and variance of the output of each layer for anomalies. Add integration tests: Run end-to-end tests to ensure the logical integrity of model training, generated data. Compare the outputs of different models and evaluate the generative effect of CTGAN with other deep learning models. Perform downstream task tests using synthetic data to validate data quality