-
TODO experiment with fft prediction
-
tensorflow python version of a bitcoin trader - https://github.com/lefnire/tforce_btc_trader
- excelent intro to the field... with even teaching on trading!!!
- podcast associated to it - http://ocdevel.com/mlg/26
-
https://stackoverflow.com/questions/39138447/how-to-stack-lstm-layers-using-tensorflow?rq=1
-
basic example of lstm + tensorflow.js
-
lstm example: https://github.com/tensorflow/tfjs-examples/blob/master/addition-rnn/index.js
-
Do web, no node.js
- still run in the browser + es6 module
- this is the environment you know
-
about lstm parameters - https://stackoverflow.com/questions/49573242/what-is-sequence-length-in-lstm
-
timeseries with only dense layer ??? see it here
- https://medium.com/@andrewtrahan/trials-and-triumphs-of-time-series-machine-learning-with-tensorflow-js-4f1a1140fe10
- it could temporarily remove the issue with the lstm setup... and unblock
-
Very complete implementation
-
now that you got a kind of prediction, display the graph
- delta between prediction and label - how to visualise that
- what if only prediction based on a single time window
- https://thecodingtrain.com/CodingChallenges/100.1-neuroevolution-flappy-bird.html
- https://blog.apptension.com/2018/06/27/tensorflow-js-machine-learning-and-flappy-bird-frontend-artificial-intelligence/
- make a lstm run in the browser
- try to find a time series
- no over engineering at first
- run the lstm on csv
- display the result
- may be nice in tfvis
- DONE download a stock in csv
- DONE display csv in canvas