MSc Thesis: Mapping the Meander: A Comprehensive Review and Benchmarking of Automatic Mind Wandering Detection
12.07.2024, Abschlussarbeiten, Bachelor- und Masterarbeiten
Given the novelty and relevance of the topic, the review has strong potential for publication in high-impact journals related to cognitive science, machine learning, and human-computer interaction.
Objectives
Literature Search
A thorough literature search will be conducted using academic databases such as IEEE Xplore, PubMed, and Google Scholar. Keywords will include "mind wandering detection," "automatic detection of mind wandering," "cognitive drift detection," and "benchmarking in mind wandering."
Comprehensive Dataset Analysis
The review will catalog and analyze the datasets commonly employed in mind wandering detection studies. This includes public datasets as well as notable private ones, detailing their characteristics, sample sizes, data collection methods, and features.
Model Benchmarking
The review will evaluate the most prevalent machine learning and deep learning models used for different modalities in mind wandering detection. This involves an in-depth analysis of model architectures, feature extraction techniques, and performance metrics.
Performance Comparison
Through benchmarking, the review will compare the performance of these models across different datasets. This will highlight which models perform consistently well and identify potential gaps in the current research landscape.
Methodological Insights
The review will identify best practices and common pitfalls in the field, providing insights into effective techniques for data preprocessing, model training, and evaluation.
Expected Outcomes
- Comprehensive Review
- Benchmarking Report
- Publication Potential
Timeline
Month | Aim |
---|---|
1 | Finding most suitable search criteria Systematic review: title and abstract screening |
2 | Paper reading Data preprocessing of previous datasets |
3-5 | Writing the introduction and related work of the thesis Writing the methodology Running experiments, benchmarking |
6 | Summarizing the results Writing the experiments, results, and discussion |
Weekly update meetings would take place in Marsstr. 20-22 or online.
Requirements
- Programming knowledge (Python, Pytorch)
- Machine learning theoretical and practical knowledge
Advisors and Supervisor
Advisors: Anna Bodonhelyi
Supervisor: Prof. Dr. Enkelejda Kasneci
Important Dates
Start date: July
Thesis registration: middle of September (based on progress)
Are you interested? Send your CV and transcript of records to anna.bodonhelyi@tum.de to arrange an interview!
Kontakt: anna.bodonhelyi@tum.de