Project Link
About
Reinforcement learning is used for prediction.
In this technology, you can create an environment (sort of like a gym) to train your robot. Depending on the environment and the reward for the action, our simulated robot gains the ability to learn. Created a virtual gym with a lot of instrumentation data. Initially, the robot has an initial balance, based on experience, it decides what to do, buy, sell or do nothing.
The author claims to have achieved amazing results, and the robot was able to increase its original portfolio by 50%: "To be honest, this is a very fun experience, like a rabbit hole."
The learning process can be observed at this link: http://www.keeppredict.com:6006/
And the neural network itself is here: http://www.keeppredict.com:6006/#graphs&run=Keep_learn_1
Github:
Future plans:
In this technology, you can create an environment (sort of like a gym) to train your robot. Depending on the environment and the reward for the action, our simulated robot gains the ability to learn. Created a virtual gym with a lot of instrumentation data. Initially, the robot has an initial balance, based on experience, it decides what to do, buy, sell or do nothing.
The author claims to have achieved amazing results, and the robot was able to increase its original portfolio by 50%: "To be honest, this is a very fun experience, like a rabbit hole."
The learning process can be observed at this link: http://www.keeppredict.com:6006/
And the neural network itself is here: http://www.keeppredict.com:6006/#graphs&run=Keep_learn_1
Github:
- the code can be found here https://github.com/airocket/keep_dl
- final results https://github.com/airocket/keep_rl
Future plans:
- collect more data,
- create more accurate models,
- adjust hyperparameters,
- add visual interface (website)