Examples and code used in the presentation "Deep learning in Python?" at Meetup Deep Learning Ljubljana on 2016-06-20.
We will dive into popular Python deep learning frameworks such as Theano, TensorFlow, and Keras and demonstrate a few hands-on examples from classic machine learning, computer vision, and natural language processing. After comparing the resulting code of dense, convolutional (CNN), and recurrent neural networks (RNN), we will take a glimpse at how Keras works under the hood. Additionally we will see how to kick-start your deep learning research and move to production using Docker containers.
Requirements:
- Docker to use already prepared container images (
gw000/keras-full
orgw000/keras
) - (or alternatively an environment with Keras, Theano, and TensorFlow)
Using Docker container gw000/keras-full
with Keras, Theano, TensorFlow, Python 2 and 3, and the Jupyter Notebook web interface (http://localhost:8888/) on CPU:
$ docker run -it -p=8888:8888 gw000/keras-full
Using Docker container gw000/keras-full
on GPU with notebooks in /srv/notebooks/
:
$ docker run -it $(ls /dev/nvidia* | xargs -I{} echo '--device={}') -p=8888:8888 -v=/srv/notebooks:/srv gw000/keras-full
Copyright © 2016 gw0 [http://gw.tnode.com/] <gw.2016@tnode.com>
Copyright © 2015 François Chollet, Google, Inc. and respective contributors to some examples in Keras
Slovenian poetry and prose is from http://lit.ijs.si/. I.D.I.O.T poetry and prose is from http://id.iot.si/.
This code is licensed under the MIT license (MIT).