Adi is a Lead Data Scientist at PayPal. She develops machine learning models for fraud detection, which are at the core of PayPal’s risk system. Adi is particularly excited about leveraging XAI (eXplainable AI) to improve the ML models’ performances and their social impact. Adi holds a BSc and an MSc in Computer Science from the Hebrew University and the Weizmann Institute, respectively. She also volunteers on the Public Knowledge Workshop’s Open Bus team, open-sourcing Israel’s real-time public transportation data.
Adi is a Lead Data Scientist at PayPal. She develops machine learning models for fraud detection, which are at the core of PayPal’s risk system. Adi is particularly excited about leveraging XAI (eXplainable AI) to improve the ML models’ performances and their social impact. Adi holds a BSc and an MSc in Computer Science from the Hebrew University and the Weizmann Institute, respectively. She also volunteers on the Public Knowledge Workshop’s Open Bus team, open-sourcing Israel’s real-time public transportation data.
As data scientists, we develop models that take action and affect people’s lives. These models are not perfect – even high-performing models are challenged by our ever-changing reality. In such dynamic circumstances, effective methodologies for debugging ML models are extremely important for a data scientist looking to build impactful, robust solutions. Of course, as with any real-world task, there is no single, perfect solution, but we can and should understand what tools are at our disposal and can be employed for our purposes. We will start the discussion with different scenarios where model debugging is required, what we can gain from such a process, and how it can be evaluated. We will explore several model debugging approaches, share our own experiences, and learn from others sharing theirs
As data scientists, we develop models that take action and affect people’s lives. These models are not perfect – even high-performing models are challenged by our ever-changing reality. In such dynamic circumstances, effective methodologies for debugging ML models are extremely important for a data scientist looking to build impactful, robust solutions. Of course, as with any real-world task, there is no single, perfect solution, but we can and should understand what tools are at our disposal and can be employed for our purposes.
We will start the discussion with different scenarios where model debugging is required, what we can gain from such a process, and how it can be evaluated. We will explore several model debugging approaches, share our own experiences, and learn from others sharing theirs.
© 2018-2021 WiDS TLV – Intuit. All rights reserved.
Scotty – By Nir Azoulay
Design: Sharon Geva