- Deep Learning Picture ("drilling down") (30 sec)
- Content (30 sec)
- Computational Paradigms (3 min)
- Explain computational graphs in one slide
- Draw a TensorFlow comp. graph
- Explain Operation Types
- Variables
- Explain Tensors in terms of symbolic handle and in terms of edges
- Explain Sessions, especially
runand how graphs are executed
- Execution Model (2 min)
- Why is Execution + Hardware so important (it really is!)
- Client, Master, Workers, Devices
- Two Degrees of Scalability (many-device vs many-machine)
- Placement Algorithm (why TensorFlow supports these two degrees so well)
- Backpropagation (2 min)
- Visualization with TensorBoard (2 min)
- Why necessary
- Explain with online Demo
- Graph
- Expansion of Scopes
- Scalar Summaries
- Histogram Summaries
- Images
- Use cases of TensorFlow today (2 min)
- Email Examples (30 sec)
- Drug Discovery (30 sec)
- DeepMind (30 sec)
- Walkthrough (5 min)