University of Twente Student Theses
Explaining Hiccup Predictor using Layerwise Relevance Propagation
Neupane, Nischal (2023) Explaining Hiccup Predictor using Layerwise Relevance Propagation.
Full text not available from this repository.
| Full Text Status: | Access to this publication is restricted |
| Embargo date: | 29 November 2028 |
| Abstract: | With increased complexity in systems and architectures, using Natural Language Processing (NLP) models for analysis tasks becomes crucial for organizations like ASML. Optimizing and analyzing event logs and operational processes is vital for ASML systems to run consistently and reliably. The thesis provides insights and an understanding of the differences and similarities between various explainability methods for Root Cause Analysis (RCA) for ASML's event log analysis and process optimization. Opening up black-box models is essential, mainly when applied in sensitive domains, such as optimizing manufacturing processes at ASML. A growing portion of eXplainable AI (XAI) research has successfully used explainability methods to auxiliary output explanations at inference time together with predictions made by these complex models. When scoping the different methods, there is a distinction to be made regarding whether the explanations emerge as part of the prediction process or subsequently via a separate model. These two main categories of explainability methods are self-explaining and post-hoc, respectively. This work focuses on the post-hoc evaluation, analysis, and comparison approach through the use of Layer-wise Relevance Propagation. Through systematic experimentation, we identified a window size of 1,024 as optimal for enhancing interpretability and the nuanced relationship between model complexity and data characteristics in influencing RCA outcomes. The overall outcomes resulted in vastly inconclusive results where further exploration is required. These findings, while specific to ASML's challenges, offer broader implications, enriching the overarching discourse on data science and model explainability. |
| Item Type: | Essay (Master) |
| Clients: | ASML, Veldhoven, Netherlands |
| Faculty: | EEMCS: Electrical Engineering, Mathematics and Computer Science |
| Subject: | 54 computer science |
| Programme: | Computer Science MSc (60300) |
| Link to this item: | https://purl.utwente.nl/essays/97710 |
| Export this item as: | BibTeX EndNote HTML Citation Reference Manager |
Show download statistics for this publication
Show download statistics for this publicationRepository Staff Only: item control page