Trustworthy AI Seminar at UPV

🌱 I’ve been invited by the School of Agricultural Engineering and Environment (ETSEAMN) at Universitat Politècnica de València (UPV) to deliver a seminar as part of their “Artificial Intelligence and Emerging Technologies” seminars.

The class will explore Trustworthy AI from the perspectives of interpretability and uncertainty quantification, and it’ll be divided into three main parts:

• Part I: While interpretable AI covers a broad spectrum of topics, we’ll focus on the fascinating area of equation discovery, also known as Symbolic Regression (SR). This field plays an important role in scientific discovery and interpretable modeling. We’ll look at a decomposable SR method that distills a trained “opaque” regression model into mathematical expressions, which serve as interpretations of its computed function.

• Part II: After learning about SR, it’s important to recognize that the accurate recovery of symbolic representations may be influenced by different types of uncertainty. Understanding and quantifying them is essential, especially in application domains where ensuring the reliability of AI-powered systems is critical.

• Part III: Finally, we’ll explore a practical application in precision agriculture. Here, we’ll discuss how to learn site-specific mathematical expressions that relate fertilizer rate and expected crop yield.

The slides, including the Google Colab links, are available below: