Edoardo Mosca


Room: FMI 01.05.057
E-Mail: edoardo.mosca(AT)tum.de
Phone: +49-89-289-18684
Address: TUM - Fakultät für Informatik, Boltzmannstr. 3, 85748 Garching

Research Interests

I work in the field of eXplainable AI (XAI) for Machine Learning and Deep Learning with a strong focus on NLP. In other words, besides building accurate machine learning models, my goal is also to understand why those models make certain decisions. 

For instance, these are some of my research questions:

  • If we are given a black-box model that solves a problem well, can we explain its decision without sacrificing predictive performance?
  • Can we, though XAI techniques, understand (and hence avoid) strange and detrimental phenomenons such as uninted bias and vulnerability to adversarial examples? 
  • Human-in-the-loop ML: Can we use human rationale to feed explanations back into a model? Can this process give back to humans influencial power over the decision-making process? 

Here you can find my master thesis about XAI techniques applied in the field of Hate Speech Detection. There is also a blog post article, partially inspired by my thesis, that briefly discusses why XAI is so relevant in modern machine learning research and explores a possible definition of XAI (thanks to Cingis Alexander).


Each semester, in collaboration with some of my colleagues, we offer advanced master-level practical courses for students motivated to get hands-on experience in XAI, NLP, and ML. In this courses you will team up with other participants and work on a concrete project that utilized or even tries to improve the current state of the art in the field. The projects last for the whole semester and will be under my close supervision. We will meet on a (bi-)weekly basis to discuss the current state and the next steps. At the end of the semester, you will be presenting your findings and results as well as delivering your code and a brief "paper-like" report.

Each Winter Semester:
Master Lab Course - Explainable AI for Machine Learning (IN2016, IN4286)

(10 ECTS) see on TUMOnline

Each Summer Semester:
Master Lab Course - Machine Learning for Natural Language Processing Applications (IN2016, IN4286) 
(10 ECTS) see on TUMOnline

If you're interested, it is always a good idea to contact the advisor/manager of your study program. Ask them if the courses are recognized for your master degree.


To my Google Scholar profile

Mosca, E., Wich, M., & Groh, G. (2021, June). Understanding and Interpreting the Impact of User Context in Hate Speech Detection. In Proceedings of the Ninth International Workshop on Natural Language Processing for Social Media (pp. 91-102).

Wich, M., Mosca, E., Gorniak, A., Hingerl, J., & Groh, G. (2021, September). Explainable Abusive Language Classification Leveraging User and Network Data. In Joint European Conference on Machine Learning and Knowledge Discovery in Databases (pp. 481-496). Springer, Cham.

Previous to my PhD:

Servadei, L., Mosca, E., Zennaro, E., Devarajegowda, K., Werner, M., Ecker, W., & Wille, R. (2020). Accurate cost estimation of memory systems utilizing machine learning and solutions from computer vision for design automation. IEEE Transactions on Computers69(6), 856-867.

Servadei, L., Mosca, E., Devarajegowda, K., Werner, M., Ecker, W., & Wille, R. (2020, September). Cost estimation for configurable model-driven SoC designs using machine learning. In Proceedings of the 2020 on Great Lakes Symposium on VLSI (pp. 405-410).

Servadei, L.*, Mosca, E*., Werner, M., Esen, V., Wille, R., & Ecker, W. (2019, September). Combining evolutionary algorithms and deep learning for hardware/software interface optimization. In 2019 ACM/IEEE 1st Workshop on Machine Learning for CAD (MLCAD) (pp. 1-6). IEEE. (* equal contribution)