anything that implements ARIMA

There are lots of engineering tricks given fancy names in ML. No point in memorizing all of them, but keep reading them to maintain intuition.

Probably because most aren’t statistically significant…

If the model has the following format:

Encoder, diffusion in latent space, decoder

you can average the embedding right before the decoder. Just make sure your averaging is of a representation that can be meaningfully averaged. E.g. averaging images would result in garbage, but their corresponding embeddings might provide something useful.

No way that’s exactly what I’ve been doing as a side project.