Efficient Projection onto the Perfect Phylogeny Model

This code implements a generalization of the algorithm described in the paper "Efficient Projection onto the Perfect Phylogeny Model".

Please cite this code as:

@inproceedings{projection-onto-PPM, title={Efficient Projection onto the Perfect Phylogeny Model}, author={Jia, Bei and Ray, Surjyendu and Safavi, Sam and Bento, Jose}, booktitle={Advances in Neural Information Processing Systems}, year={2018} }

---

Instructions:

• To compile the code just do

gcc project_onto_PPM.c 

• To run the code just do

./a.out InputDataExample OutputData 0    

or

./a.out InputDataExample OutputData 1

• The input data is of the following form, see the example file for exact formating instructions.

  • number of nodes
  • number of samples
  • matrix of mutation frequencies, Fhat (column major form)
  • vector that scales the norm in the objective
  • root node
  • vector of degree of each node in the tree
  • adjancency list (neighbors of each node in the tree)
  • flag, 0 or 1, that indicates whether to output the inferred fraction of each muutant, M, or just the projection cost, C(U)

• When calling the program, the third argument, 0 or 1, indicates whether the columns of M must sum to 1, or sum to something smaller than 1.

• If you want to test the code on random inputs, and compare against the output produced by CVX-Matlab, you can run the script test_against_CVX_matlab.m in Matlab, after you have installed CVX http://cvxr.com/cvx/download/

• The dependencies of the different functions in the code is described by the following diagram.