What it does This is the archived repository for “Creative Applications of Deep Learning w/ TensorFlow,” a three-course MOOC that ran on Kadenze Academy. It contains lecture transcripts and homework assignments as Jupyter notebooks, plus a companion Python package (pycadl). The curriculum moves from basic TensorFlow operations through autoencoders, Deep Dream, GANs, and into Magenta, WaveNet, and PixelCNN.

The interesting bit The syllabus itself is a time capsule of 2016–2017 deep learning pedagogy: teaching raw TensorFlow graph construction before Keras became the API of choice, and treating notebooks as the primary submission format for coursework. The Docker and pip install instructions are extensive because getting GPU support working was still genuinely painful.

Key highlights

• Three courses, ~13 sessions total, covering CNNs, RNNs, GANs, VAEs, and autoregressive models
• Each session pairs a lecture notebook with a homework notebook
• Includes Docker and nvidia-docker setup for reproducible environments
• Companion package pycadl maintained separately for courses 2 and 3
• Some notebooks have Colab links for running without local setup

Caveats

• Repository is explicitly archived; materials are historical, not maintained
• TensorFlow 1.x-style code throughout (the README references pip install tensorflow-gpu)
• Windows GPU support was limited even then; Docker Toolbox was the recommended workaround

Verdict Worth browsing if you’re researching how deep learning was taught circa 2016–2017, or if you want to see early implementations of Deep Dream and style transfer in raw TensorFlow. Skip it if you’re looking for current, runnable tutorials — the framework versions and APIs have moved on.