Mitchell's AI Workshop

Artificial Intelligence

On this page... (hide)

1. Preliminaries
2. Training Data, Evaluation Data, Performance Metrics
3. CNN as a Greybox
4. CNN Tutorials
5. Diving Deeper into CNN

1. Preliminaries

Have a laptop, with text editor
Install python 2.7 and 3.6 in each separate Conda environments
- Favourite python learning tutorial Tutorialspoint
  - Only spend max 2 hours on your own
- Conda will manage versioning and protection of all applications and packages
  - These will persist and still be there even if you return some time later, all will still work
  - Link to Conda tutorial MiniConda

2. Training Data, Evaluation Data, Performance Metrics

Data manipulation, presentation in python

3. CNN as a Greybox

Feature Extraction, Cost Function, Minimisation in a tensor context, Neural Net, Nodes, Weights
- Figure 1.2 of Deep learning Book Deep Learning
- Browse slides and notes from the Indaba item on Mathematics for Machine Learning slides

4. CNN Tutorials

Jump straight in to Keras and Tensorflow examples like in the QuickStart link
- Test python program that creates a CNN to classify digits in the MNIST data set
- Running on your GPU, Amazon cloud, Google engine

5. Diving Deeper into CNN