f-INE: A Hypothesis Testing Framework for Estimating Influence under Training Randomness

1. 1Identify the need for influence estimation in machine learning models
2. 2Recognize the limitations of existing methods due to training randomness
3. 3Apply f-INE framework to estimate influence under training randomness
4. 4Interpret the results to inform data curation and cleanup decisions

Machine learning engineers and researchers benefit from f-INE as it helps to explain and debug models by estimating the impact of individual samples on the final model, which is crucial for data curation and cleanup

💡 f-INE provides a reliable way to estimate the impact of individual samples on the final model despite training randomness

🤖 Introducing f-INE: a framework for estimating influence in ML under training randomness