Conclusion

The take home points are:
1. Bump attractors can handle continous phenomena (old).
2. There are problems with large ranges (new), that can be overcome.
3. In this BBC AM, learning makes for a reasonable cognitive model.
4. This has been integrated into the NEAL framework if you want to use them for agents with neurons.
5. We really need better spiking models of CAs.
There's more on https://www.cwa.mdx.ac.uk/NEAL/wta.html.
Huyck, C. R., & Vergani, A. A. (2020). Hot coffee: associative memory with bump attractor cell assemblies of spiking neurons. Journal of Computational Neuroscience, 48(3), 299-316.
Vergani, A. A., & Huyck, C. R. (2020). Critical limits in a bump attractor network of spiking neurons. arXiv preprint arXiv:2003.13365, and a NICE-2020 poster.