This attachment contains the following:

- data
    - ad_app_test_data: Anomaly data used to test Anomaly Detection.
    - pm_lab_data: Measurements and generated anomaly data 
- src
    - impl
        - mocker: Python point machine mocker.
        - keras: Jupyter notebooks of CAE models in Keras and Bayes optimizations.
    - thesis: Latex source files of the thesis text.
- video
    - flow_creator: Recordings of the turnout app replica done in Flow Creator.
    - honu-ae: Recordings of the HONU-AE model.
    - video/100ms_dta: Recordings of the 100 DTA model.
- text: PDFs of the thesis text.


Anomaly files follow this convention:
- anom_peaks: Peak anomalies inserted after the initial acceleration phase.
- anom_sin_add_x5: Anomalies where a sine wave perturbation is added to the original signal. The added sine wave is multiplied by 5.
- anom_sin_add_x20: Anomalies where a sine wave perturbation is added to the original signal. The added sine wave is multiplied by 20.
- anom_sin_replace: Anomalies where a shapelet sine wave replaces a part of the original signal.

* Note about the HONU-AE and DTA recordings:
The videos show the iterative version of the HONU model, so the videos were paused while waiting for it to finish processing data.
Some information was blurred to protect sensitive information.
The application was taking longer than usual to send the results in the recording showing a left rotation on the DTA 100ms model, hence the looking around in IIH Essentials and Anomaly Detection at cca 00:22-00:39. This was most likely a result of running too many resource-intensive applications alongside the IED and IEM.