Affinity propagation

1.1 Reading
As first step, familiarise yourself with affinity propagation, for example by working through
https://en.wikipedia.org/wiki/Affinity_propagation.
1.2 Task 1: Prepare a test dataset
Pick a small set of values that enable you to apply clustering using affinity propagation on paper.
Marking information: Up to 10 points: clarify why you believe your small dataset to be
suitable for clustering.
1.3 Task 2: Apply affinity propagation on paper
Work through your small dataset from Task 1 to ensure you have known-good results.
Marking information: Up to 20 points: include information on how you sanity-checked
your results as the number of iterations may considerably affect the precision of your
final values.
1.4 Task 3: Implement affinity propagation in Python
You are now set to actually add the implementation. Note that you are expected to fully implement
the mathematical operations instead of using a library function such as scikit or statsmodels.
Your implementation may make several assumptions about the inputs provided to it. Make those
1
explicit in comments. Also, provide an estimate on how long execution of your algorithm will take
dependent on the input values. Express this estimate as a function of the inputs, such as the number
of clusters and/or the number of data points.
Marking information: Up to 50 points: 30 points for a correctly working Python implementation, 10 points for describing limitations and assumptions of your implementation,
and 10 points for a description of the complexity of your algorithm.
1.5 Task 4: Create a test harness
Test your implementation via unit and integration tests. As part of this work, you may choose to
compare to a reference implementation, such as sci-kit learn’s.
Marking information: Up to 20 points: identify suitable unit tests and integration tests.
Explain what coverage you expect your test suite to have.