Kellian Cottart

Writing

Long-form pieces: the mechanisms behind Bayesian and continual learning, worked through.

2026

  1. 21 Sep Essay

    When Bayesian neural networks remember too much

    A Bayesian posterior that accumulates every observation ends up too certain to learn anything new.

  2. 4 May Essay

    A geometrical interpretation of variational methods in machine learning

    Variational methods are often introduced through information theory: learning means finding an approximate distribution that remains close to a target Bayesian posterior.

2 posts since 2024. New ones arrive in the RSS feed.