jfaleiro/deeplearning repository overview

⁠Deeplearning

Deep-learning in a box. Simple, one liner. An image with all must-have deep-learning tools and a sample of time series prediction using a recurrent neural network.

License⁠

One image host complete with PyNum, Pandas, Tensorflow, Tensorboard, and all the other must-have deep-learning tools. To activate it, one command, one line:

docker run -p 8888:8888 -p 6006:6006 -v `pwd`:/work --name dl deeplearning

This of course assumes you have docker installed with sudo privileges. The first time you run this command the latest image will be downloaded from my docker hub⁠.

To use it, after activation, just point your browser to one of the ports 8888 for your Jupyter console:

...and 6006 for your Tensorboard dashboard:

The image carries a toy example of a time series prediction model using RNN. You can browse through the example in the notebook timeseries_RNN.ipynb.

After you execute it once, you will be able to click through the dashboard to visualize the RNN Tensorflow graph:

The execution also generates the convergence plot of the loss function, decaying asymptotically to near zero in a span of a few seconds:

You can adjust the model for prediction of other time series, or create new ones, leveraging the same framework.

Enjoy!