Matthew J. A. Smith

2 accepted papers

2026

Influence Dynamics and Stagewise Data Attribution

Jin Hwa Lee, Matthew J. A. Smith, Maxwell Adam, Jesse Hoogland

ICLR 2026poster

Current training data attribution (TDA) methods treat influence as static, ignoring the fact that neural networks learn in distinct stages. This stagewise development, driven by phase transitions on a degenerate loss landscape, means a sample's importance is not fixed but changes throughout training…

Cited by 0 Source ↗Scholar ↗

2023

Why Target Networks Stabilise Temporal Difference Methods

Mattie Fellows, Matthew J. A. Smith, Shimon Whiteson

ICML 2023poster

Integral to recent successes in deep reinforcement learning has been a class of temporal difference methods that use infrequently updated target values for policy evaluation in a Markov Decision Process. Yet a complete theoretical explanation for the effectiveness of target networks remains elusive.…

Cited by 11 Source ↗PDF ↗Scholar ↗