Hacking on Creative Prediction

Here’s some exercises/projects/tutorials, to get you started with creative prediction.

Remember to check out the setup page to get your Python/Keras/Jupyter tools running.

1. Set up Python, Keras, and Jupyter

2. Try out a Colab Notebook

3. Make your own text generator

4. Make some music with an RNN

5. Make a RoboJam model

6. Try out IMPS, Processing, and Pd

7. Design a Predictive NIME

Tutorials/Demos

• Generating Creative Sequences

  • Generating Text with a CharRNN; inventing Star Trek episode titles. github colab

  • Advanced CharRNN techniques. github colab

  • Continuing musical sequences. github colab

  • Combining multiple sequences in a CharRNN with “Gesture-RNN” (not working right now..). github colab

• RNNs for continuous-valued data: Mixture Density Layers (MDNs)

  • Mixture distribution examples github colab

  • Introduction to Mixture Density Layers github colab

  • Predicting sketching: Kanji generation with a Mixture Density RNN. github colab

  • Predicting time and place - musical scribbles with RoboJam. github colab

• VAE, GAN, and World Models

  • DoomVAE example github colab

  • Doom World Model example (not done yet!)
  • MicroGAN example github colab