Chapter 7

This section presents how the results presented in Chapter 7 can be reproduced. The corresponding scripts are all located in the pdm folder.

1. Parsing

Unfortunately, this proprietary dataset is subject to strict confidentiality levels and therefore cannot be published.

The sampling frequency that the dataset is downsampled to is obtained the same way as for the PATH dataset in Chapter 4. The corresponding code is found in the pdm_downsampling.py script.

The results were obtained as follows. First, the raw MF4 files are parsed using the 0_parse.py script.

2. Preprocessing

Then, preprocess the contents of the 1_parsed folder using 1_data.py, which includes downsampling, standardisation, and windowing.

3. Training

Once preprocessed, the multi-model TeVAE model training can be run with the 2_training.py script.

4. Inference

Before evaluating the trained multi-model TeVAE model, it needs to be run on the validation and test subsets, which is done with the 3_inference.py script. Again, AD_MODE and MODEL_NAME carry the same function as in the scripts preceding it.

5. Evaluation

With the inference outputs, the results can now be evaluated using the 4_evaluation.py script. Again, AD_MODE and MODEL_NAME carry the same function as in the scripts preceding it.