Timezone: »

Online Structured Laplace Approximations for Overcoming Catastrophic Forgetting
Hippolyt Ritter · Aleksandar Botev · David Barber

Thu Dec 06 02:00 PM -- 04:00 PM (PST) @ Room 517 AB #135

We introduce the Kronecker factored online Laplace approximation for overcoming catastrophic forgetting in neural networks. The method is grounded in a Bayesian online learning framework, where we recursively approximate the posterior after every task with a Gaussian, leading to a quadratic penalty on changes to the weights. The Laplace approximation requires calculating the Hessian around a mode, which is typically intractable for modern architectures. In order to make our method scalable, we leverage recent block-diagonal Kronecker factored approximations to the curvature. Our algorithm achieves over 90% test accuracy across a sequence of 50 instantiations of the permuted MNIST dataset, substantially outperforming related methods for overcoming catastrophic forgetting.

Author Information

Hippolyt Ritter (University College London)
Aleksandar Botev (University College London)
David Barber (University College London)

More from the Same Authors