timothyyu / gdax-orderbook-ml

Application of machine learning to the Coinbase (GDAX) orderbook
BSD 3-Clause "New" or "Revised" License
95 stars 28 forks source link

test implementation of keras.js for reduced complexity model #41

Open timothyyu opened 6 years ago

timothyyu commented 6 years ago

https://github.com/transcranial/keras-js

this plus mlive/heroku + or a combination of dedicated or virtual VPS instance due to model bandwith and data size requirements (current)

lorrp1 commented 6 years ago

Woud be nice a reinforcement learning bot with this prediction based on s/r

timothyyu commented 6 years ago

Semi/psuedo supervised learning + reinforcement learning would be ideal, but there is a list of issues that have to be addressed first (even before the hosted model stage) to get to that point, including:

31 - Gutting all matplotlib.finance dependencies

26, #25, #22, #20 - model specific abstractions/improvements + reduction in memory requirements

9, #10, #21 - data scraping/async library component integration/revision

timothyyu commented 6 years ago

potential for model to run entirely in js with node backend with keras.js