Tech Meridian ← LIVE FEED
RU

NEWS · RESEARCH · #306

Google Research: instance-conditional timescales of decay to reweight training data under concept drift

Google Research proposes learning an auxiliary, instance-conditional weighting model that assigns importance scores to training examples as a function of their content and age, combining multiple fixed timescales of decay and meta-learning the assignment alongside the primary model. The method aims to blend benefits of offline and continual learning and yields up to ~15% relative accuracy gains on a large nonstationary photo benchmark (~39M images over 10 years) and improvements across other nonstationary learning benchmarks.

KEY POINTS

  1. Google Research proposes learning an auxiliary, instance-conditional weighting model that assigns importance scores to training examples as a function of their content and age, combining multiple fixed timescales of decay and meta-learning the assignment alongside the primary model.
  2. The method aims to blend benefits of offline and continual learning and yields up to ~15% relative accuracy gains on a large nonstationary photo benchmark (~39M images over 10 years) and improvements across other nonstationary learning benchmarks.
  3. This approach offers a practical, learned way to reduce model degradation under slow concept drift by selectively downweighting or preserving past examples, addressing limitations of pure offline or continual strategies in real-world nonstationary data.

WHY IT MATTERS

This approach offers a practical, learned way to reduce model degradation under slow concept drift by selectively downweighting or preserving past examples, addressing limitations of pure offline or continual strategies in real-world nonstationary data.

SOURCES & TIMELINE

1