Single-model multi-asset workflows

[qAp]

The different cryptocurrencies are shown to have some correlation with each other, so it would seem to make sense to model them together, rather than training a model for each of them.

What are some examples of this kind of workflow?

[qAp]

Simple Transformer from Keras

• https://www.kaggle.com/yamqwe/crypto-forecasting-transformer-keras-tuner

Long-Range Transformers for Dynamic Spatiotemporal Forecasting

• https://towardsdatascience.com/multivariate-time-series-forecasting-with-transformers-384dc6ce989b
• https://arxiv.org/abs/2109.12218
• https://github.com/QData/spacetimeformer

A Transformer-based Framework for Multivariate Time Series Representation Learning

• https://www.kaggle.com/yamqwe/time-series-transformer-written-from-scratch
• https://arxiv.org/abs/2010.02803