VAE

Application of Gaskin et al.'s neural-net parameter calibration methodology to Pan et al.'s SoTA CLSNA (co-evolving latent space network with attractors) opinion dynamics model.

Additionally, reframes problem as a VAE and explores many encoder architectures including CNNs and GNNs (as part of a Variational Graph Auto Encoder architecture).

The results obtained in this repository are reproducible and can be provided at request.