What question did this study set out to answer?

The aim is to develop a method to assess the influence of training examples on model predictions.

February 19, 2020Open Access

Estimating Training Data Influence by Tracing Gradient Descent

View Full Paper

Ask AI

Bookmark

The aim is to develop a method to assess the influence of training examples on model predictions.
Introduced TracIn, which tracks loss changes for test points based on specific training examples.
Implemented a first-order gradient approximation for scalable computation.
Utilized saved checkpoints and selective layers of deep neural networks.
TracIn successfully simplified the implementation process for evaluating training data influence.
Demonstrated effectiveness across various machine learning models regardless of architecture.
Popularized the use of gradient information, checkpoints, and loss functions for assessing training data.

No takes yet. Share an insight, caveat, or question.

Randomized trial measures how training data influences predictions in machine learning models, suggesting improvements for training processes.

Pruthi et al. (2020) studied this question.

View Full Paper

Ask AI

Bookmark

Randomized trial measures how training data influences predictions in machine learning models, suggesting improvements for training processes.