[Framework|RL] Introduce non-linearity and noise in artificial data set

mjuchli / ctc-executioner

Master Thesis: Limit order placement with Reinforcement Learning

176 stars 83 forks source link

[Framework|RL] Introduce non-linearity and noise in artificial data set #18

Closed mjuchli closed 6 years ago

mjuchli commented 6 years ago

Currently the artificial data set configurator only allows to create linear price trends (up/down/flat). We should also be able to simulate a

[x] 1. price trend according to some function f (e.g. sin, or a step function).
[ ] 2. and introduce some sort of noise

With that we can show that the model is capable of approximating more complex functions (1) and can generalize on the main pattern (2).

mjuchli commented 6 years ago

For now there is no need for extra noise within the order book states. A non-linear pricing function already seems to introduce difficulties for the learner to generalize.