EURO-Online login
- New to EURO? Create an account
- I forgot my username and/or my password.
- Help with cookies
(important for IE8 users)
265. Optimizing machine learning: Enhancing interpretability and performance through mathematical optimization
Invited abstract in session TD-1: Veronica Piccialli, stream Keynotes.
Tuesday, 14:30-16:00Room: Sportshallen (building: 101)
Authors (first author is the speaker)
1. | Veronica Piccialli
|
DIAG, Università degli Studi di Roma "La Sapienza" |
Abstract
This presentation highlights the interplay between optimization and machine learning, examining two pivotal examples.
First, we will explore how mathematical optimization can significantly enhance the interpretability of machine learning models, with a focus on classification problems. Interpretability is a crucial requirement for the practical deployment of machine learning models across diverse fields. Optimal decision trees are fundamental tools for building robust and interpretable classification models. However, the challenge lies in their inability to ensure optimality for large datasets. Recently some efficient approaches have been introduced for building optimal decision trees that require all the features to be binary. We will describe an efficient optimization-based approach for deriving a supervised discretization of the original dataset. Thanks to the use of optimization, the proposed technique allows to control the importance of the features extracted, and hence the granularity of the discretization. This novel technique not only facilitates the training of a robust and interpretable optimal decision tree but also allows to address the complexities associated with larger datasets.
Finally, we'll transition to a pragmatic application in process engineering, showcasing the successful integration of optimization and machine learning to solve an important real-world problem.
Keywords
- Machine Learning
- Optimization Modeling
Status: accepted
Back to the list of papers