Dynamic Influence Tracker: Measuring Time-Varying Sample Influence During Training
This addresses the limitation of static influence measurements for researchers and practitioners in machine learning, offering a more nuanced tool for understanding training dynamics, though it is incremental in extending existing influence tracking methods.
The paper tackled the problem of measuring how training sample influence changes over time during model training, proposing Dynamic Influence Tracker (DIT) to capture these dynamics, which achieved up to 0.99 correlation with ground truth and over 98% accuracy in detecting corrupted samples.
Existing methods for measuring training sample influence on models only provide static, overall measurements, overlooking how sample influence changes during training. We propose Dynamic Influence Tracker (DIT), which captures the time-varying sample influence across arbitrary time windows during training. DIT offers three key insights: 1) Samples show different time-varying influence patterns, with some samples important in the early training stage while others become important later. 2) Sample influences show a weak correlation between early and late stages, demonstrating that the model undergoes distinct learning phases with shifting priorities. 3) Analyzing influence during the convergence period provides more efficient and accurate detection of corrupted samples than full-training analysis. Supported by theoretical guarantees without assuming loss convexity or model convergence, DIT significantly outperforms existing methods, achieving up to 0.99 correlation with ground truth and above 98\% accuracy in detecting corrupted samples in complex architectures.