video from “plug and play generative networks: Conditional Iterative Generation of Images in Latent Space”

morphing faces online demo

a super-interesting article from google’s magenta working group on using neural networks to synthesize (musical) audio: https://magenta.tensorflow.org/nsynth-instrument

and this maybe even more interesting article (and corresponding twitter-bot about traversing latent space in images (see how to remove a smile from a face using SLERP (spherical linear interpolation), for instance)

using magenta’s port of sketch-rnn (in jupyter notebook)

explores a process of “teaching machines to draw” described here

install magenta environment (a toolkit built on top of tensorflow, specifically for “creative” tasks) following the instructions here

if you’ve already set up a cloud compute instance that includes anaconda, you can follow these steps:

type curl https://raw.githubusercontent.com/tensorflow/magenta/master/magenta/tools/magenta-install.sh > /tmp/magenta-install.sh bash /tmp/magenta-install.sh into terminal at the prompt

conda create -n magenta python=2.7 jupyter source activate magenta

pip install magenta

alternatively, you can launch the magenta docker container

if you are interested, read this thorough high-level talk by memo atken on the importance of dimensional spaces in machine learning: