Continual learning should adapt to noisy data, not over-clean it

Shahules786 · x · 2026-08-24

The dominant approach to continual learning makes data friendly to algorithms, but a more ambitious path designs algorithms that learn continuously from noisy, non-stationary experience. We may be over-investing in cleaning data for SGD.

Original post →

More from AGI Musings

AGI Musings channel →