Cntk variational auto encoder
Another implementation of variational auto encoder based on https://towardsdatascience.com/intuitively-understanding-variational-autoencoders-1bfe67eb5daf.
This variational auto encoder is trained on crypto currencies history dataset, this dataset was downloaded on https://www.binance.com with python API https://github.com/sammchardy/python-binance.
Binance proposed 129 currencies, and i took 7 variables for each currency. That made the model input vector to be 129 * 7 = 903
This a piece of the results.
| The input | The output (generated) | The difference |
|---|---|---|
 |
 |
 |