So Nvidia showcased a hedge fund algo that was sped up 6000x. Can you imagine? After my chat with a ‘guru god’ on my private chat server, it seems using mid range GPU cards could be enough? It might not cost that much. In this conversation, many tips were offer around optimizing simulations and training a model. Once Facebook regarding this Nvidia performance included:
I was lucky enough to be at an innovation lab that had 16 DGX’s. We trained a CNN in 40 minutes with a better performance against the benchmark that took days to get to the same result. It was a thing of beauty.
https://blogs.nvidia.com/blog/2019/05/13/accelerated-backtesting-hedge-funds/FACEBOOK ACCOUNT and TWITTER. Don't worry as I don't post stupid cat videos or what I eat!