Here's a brief plan of the four sessions of the workshop. Each of these sections will include exercises based on real-world datasets. While most of the workshop depends only on scikit-learn, there are a few other requirements too. An exhaustive list of Python packages required for the workshop is as follows.
- NumPy
- SciPy
- Matplotlib
- Pandas
- scikit-learn
- tensorflow
- keras
- theano
At most a couple more cursory packages might get added to this list as I proceed with creating the material, but those should be easily installable at the venue itself, assuming that the participants have a Python distribution like Enthought Canopy or Anaconda installed.
- Inbuilt dataset loading utilities
- Introduction to the estimator object
- Basic classification & regression tasks
- Introduction to supervised and unsupervised learning
- Linear vs Nonlinear models
- Kernel Methods in Machine Learning
- Feature selection & Dimensionality Reduction
- Interpreting a trained model
- Measuring model performance
- Cross validation
- Grid and random parameter search
- Gradient descent and its variations
- Introduction to neural networks
- Building a shallow neural network
- Brief introduction to deep learning