This is the code reporsitory for the Uncertainty in Deep Learning mini project. The code is an extension from https://github.com/NixGD/variational-continual-learning Based on their implementation of VCL, we introduced extentions by:
- Reordering the task sequence to mitigate the risk of severe distribution shift.
- Assigning most similar or most dissimilar classes into task to promote performance.
- The addition of scaling factors to strike a better balance between acquiring new knowledge and maintaining previous information.
