attempt absolute insanity

[timothyyu]

https://keras.io/layers/convolutional/

to look into:
keras flatten layer, permute layer, reshape layer, repeat vector

[pure size and price, potentially position also]

take l2 states as grid, feed into convolutional 2d, then flatten/reshape, then feed into lstm/gru as input in series

lambada functionality for specialized weighting of values near s/r lines?
https://github.com/keras-team/keras/blob/master/keras/layers/core.py#L564

[timothyyu]

Parallelization of input for non-sequential model (conceptual, still testing/under development):
https://github.com/timothyyu/lstm-breaker

[timothyyu]

lambada function + attention mechanism potential
https://github.com/philipperemy/keras-attention-mechanism

[timothyyu]

related to #25, #38, #26, #22