From 3c164b9f34993c2f7f5d13191649097436ea5f2b Mon Sep 17 00:00:00 2001
From: Uwe Schmitt <uwe.schmitt@id.ethz.ch>
Date: Fri, 7 Sep 2018 13:17:40 +0200
Subject: [PATCH] first version of course layout

---
 layout.md | 175 ++++++++++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 175 insertions(+)
 create mode 100644 layout.md

diff --git a/layout.md b/layout.md
new file mode 100644
index 0000000..2821adb
--- /dev/null
+++ b/layout.md
@@ -0,0 +1,175 @@
+# 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 we participants might be disctracted during
+  coding exercises and miss 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.
-- 
GitLab