Masterthesis repository by Bart-Jan Boverhof.
This repository contains all the relevant code/files for the research project Predicting Mental Workload: A Multimodal Intermediate Fusion Deep Learning Approach. For any questions please contact [email protected]. For the internal structure of the repository to work properly, it is important that the main folder is (this folder) is set as the working directory. There is an internal co-dependence in between the 1.pipeline, 2.hpo, 3.pytorch subparts of this repository, in the sense that replication should start at 1 and proceed towards 3. The output of subpart 3.pytorch are the network predictions, which are provided in this repository, such that the replicator does not need to train all networks and conduct hpo in order to replicate the final results.
A general overview of the content and structure of this repository, in addition to the required software is provided in the following table:
| Folder | Description | Utilized software |
|---|---|---|
0.manuscript |
Written LaTeX manuscript |  |
1.pipeline |
Folder that contains scripts to read data and cut into windows of pytorch tensor type. |
  |
2.hpo |
Scripts to conduct hyperparamater optimization |  |
3.pytorch |
Definition of deep-learning models and scripts to train them and test them to obtain predictions | |
4.results |
Scripts to create resulting statistics and plots from the obtained predictions | |