Zobrazeno 1 - 2
of 2
pro vyhledávání: '"Khajwal, Basim"'
It is well-known that the reparameterisation gradient estimator, which exhibits low variance in practice, is biased for non-differentiable models. This may compromise correctness of gradient-based optimisation methods such as stochastic gradient desc
Externí odkaz:
http://arxiv.org/abs/2402.11752
We study the foundations of variational inference, which frames posterior inference as an optimisation problem, for probabilistic programming. The dominant approach for optimisation in practice is stochastic gradient descent. In particular, a variant
Externí odkaz:
http://arxiv.org/abs/2301.03415