I may say this presumptiously but I always laugh when people laud AI and ML in trading. The whole key is picking your features, no matter how fancy your model is you put in garbage, you get out garbage.
-
Doesn’t seem to be true for state of the art in applications like image recognition (input is raw pixels) or generative text (input is characters) or reinforcement learning for playing games (input is raw encoding of game state)
In those domains the signal to noise ratio is much higher than in finance though so better chance of learning the important features by example alone.
-
-
Feature engineering is really just another way of exploring the bias-variance trade off by introducing stronger priors (high bias) to limit exploration of unprofitable directions in parameter space (low variance)
-
I didn't really think about it like that. Very apparent reading
@lopezdeprado 's book. Especially for games as complex as Dota, I honestly was surprised there isn't much feature engineering in openAI's model: https://d4mucfpksywv.cloudfront.net/research-covers/openai-five/network-architecture.pdf …
End of conversation
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.