This is the code repository for the reproducibility challenge coursework undertaken as part of the Advanced Machine Learning course, at the University of Oxford.
We reproduce and analyze the results of the paper Variational Continual Learning, by Nguyen et al (ICLR 2018), available freely at https://arxiv.org/abs/1710.10628.
We have reproduced the three discriminative experiments presented in the paper.
The data from our experiments can be found in TensorBoard log files in the
sub-directory final_logs/
.
We have two implementations of a Discriminative VCL model: in models.vcl_nn
, and in
models.contrib
. The former was our first implementation, has been more thoroughly
tested, and was used to obtain our experimental data; the latter is our attempt at a
cleaner implementation more similar to the PyTorch standard module style.