layout.md

# Targeted audience

- Researchers from DBIOL, BSSE and DGESS having no machine learning experience yet.
- Basic Python knowledge.
- Almost no math knowledge.

# Concepts

- smooth learning curve
- explain fundamental concepts first, discuss  exceptions, corner cases,
  pitfalls late.
- plotting / pandas / numpy first. Else participants might be fight with these
  basics during coding sessions and will be disctracted from the actual
  learning goal of an exercise.


# Course structure

## Preparation

- setup machines

- quick basics matplotlib, numpy, pandas


## Part 1: Introduction

- Why machine learning ?

- What are features ?
  - always numerical vectors
  - examples: beer, movies, images, text

- unsupervised:

  - find structure in set of features
  - beers: find groups of beer types

### Coding session:

  - read dataframe from csv or excel sheet with beer features
  - do some features vs features scatter plots
  - use tsne to show clusters
  - scikit-learn example to find clusters

## Part 2: supervised learning

- supervised:

  - classification: do I like this beer ?
    example: draw decision tree

  - classifiation: points on both sides of a line, points in circle, xor problem
    - idea of decision function: take features, produce real value, use threshold to decide
    - simple linear classifier
    - show some examples for feature engineering here to apply linear classifier

  - regression: how would I rate this movie ?
    example: use weighted sum, also example for linear regresor
    example: fit a quadratic function

### Coding session:

  - show: read circle data, plot data, augment features, learn linear classifier with scikit-learn,
    show weights and explain classifier, plot decision boundary
    load eval data set and evaluate accuracy.

  - adapt: read xor data, plot data, augment features, learn linear classifier with scikit-learn,
    show weights and explain classifier, plot decision boundary
    load eval data set and evaluate accuracy.

  - learn regressor for movie scores.


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

- how to measure accuracy ?
  - regression accuracy
  - classifier accuracy:
    - confusion matrix
    - pitfalls for unbalanced data sets
        e.g. diagnose HIV
    - precision / recall
    - ROC ?

### Coding session

- evaluate accuracy of linear beer classifier

- determine precision / recall

- ROC curve based on threshold

- provide predetermined weights, show ROC curve.


## Part 4: underfitting/overfitting

classifiers / regressors have parameters / degrees of freedom.

- underfitting:

  - linear classifier for points on a quadratic function

- overfitting:

  - features have actual noise, or not enough information
    not enough information: orchid example in 2d. elevate to 3d using another feature.
  - polnome 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
  - test data set
  - cross validation


### Coding session:

- How to do cross validation with scikit-learn
- use different beer feature set with redundant feature (+)
- run crossvalidation on classifier
- run crossvalidation on movie regression problem


## Part 5: Overview scikit-learn / algorithms

- Linear regressors
- Neighrest neighbours
- SVMs
  - demo for RBF: different parameters influence on decision line
- Random forests
- Gradient Tree Boosting
- Clustering

### Coding session

- apply SVM, Random Forests, Gradient boosting to previous examples
- apply clustering to previous examples
- MNIST example

## Part 6: pipelines / cross val / parameter optimiation with scikit-learn

- Scicit learn api
- pipelines, preprocessing (scaler, PCA)
- cross validatioon
- parameter optimization

### Coding session

- build SVM and Random forest crossval pipelines for previous examples
- use PCA in pipeline for (+) to improve performance
- find optimal SVM parameters
- find optimal pca components number

## Part 7: Best practices

- visualize features: pairwise scatter, tSNE
- PCA to undertand data
- check balance of data set, what if not ?
- start with baseline classifier / regressor
- augment data to introduce variance

## Part 8: neural networks

- overview, history
- perceptron
- multi layer
- multi layer demoe with google online tool
- where neural networks work well
- keras demo

### Coding Session

- keras reuse network and play with it.