content.md

# Targeted audience

- Researchers having no machine learning experience yet.
- Basic Python knowledge.
- Almost no math knowledge required.

# Course structure

- Two days workshop, 1.5 days workshop + .5 day working on own data / prepared data.
- Every part below includes a coding session using Jupyter notebooks.
- Coding sessions provide code frames which should be completed.
- We provide solutions.


# Day 1

## Part 0: Preparation

- Quick basics matplotlib, numpy, pandas?


### Coding session

- read dataframe from csv or excel sheet with beer features
- do some features vs features scatter plots


## Part 1: Introduction

- What is machine learning ?
- What are features / samples / feature matrix ?
- Learning problems: supervised / unsupervised


### Code walkthrough:

  - Classification: linear SVM classifier or logistic regression example
  - Clustering: scikit-learn example to find clusters.


## Part 2: classification

  Intention: demonstrate one / two simple examples of classifiers, also
             introduce the concept of decision boundary

  - Introduction: some simple two dimensional examples incl. decision function.

  - Idea of linear classifier:
    - simple linear classifier (linear SVM e.g.)
    - beer example with some weights

  - Discuss code example with logistic regression for beer data, show weights

### Coding session:

  - Change given code to use a linear SVM classifier
  - Use different data set which can not be classified well with a linear classifier


## Part 3: accuracy, F1, ROC, ...

Intention: accuracy is useful but has pitfalls

- how to measure accuracy ?

    - confusion matrix
    - accurarcy
    - pitfalls for unbalanced data sets
        e.g. diagnose HIV
    - precision / recall

### Coding session

- Evaluate accuracy of linear beer classifier from latest section
- Determine precision / recall


## Part 4: underfitting/overfitting

classifiers / regressors have parameters / degrees of freedom.

- underfitting: linear classifier on nonlinear problem

- overfitting:

  - features have actual noise, or not enough information: orchid example in 2d. elevate to 3d using another feature.
  - polynome of degree 5 to fit points on a line + noise
  - points in a circle: draw very exact boundary line

- how to check underfitting / overfitting ?

  - measure accuracy or other metric on test dataset
  - cross validation


### Coding session:

- How to do cross validation with scikit-learn
- run cross validation on classifier for beer data


## Part 5: pipelines / parameter tuning with scikit-learn

- Scikit learn API incl. summary of what we have seen up to now.
- pipelines, preprocessing (scaler, PCA)
- cross validation
- Hyper parameter tuning: grid search / random search.

### Coding session

- examples


# DAY 2

## Part 6: Overview classifiers

- Nearest neighbours
- SVMs
  - demo for RBF: different parameters influence on decision line
- Random forests
- Gradient Tree Boosting


### Coding session

- Prepare examples for 2d classification problems incl. visualization of different
  decision surfaces.

- Play with different classifiers on beer data

## Part 7: Regression

- What are differences compared to classification: output, how to measure accuracy, ...

- Example: fit polynomial, examples for underfitting and overfitting


### Coding session

Introduce movie data set, learn SVR or other regressor on this data set.


## Part 8: Introduction neural networks


- Overview of the field
- Introduction to feed forward neural networks
- Demo Keras

### Coding Session

- keras reuse network and play with it.


## Workshop

- assist to setup the workshop material on own computer.
- provide example problems if attendees don't bring own data.