Specific-and-Shared-Causal-Relation-Modeling-and-Mechanism-Based-Clustering

A unified framework for causal discovery and mechanism-based group identification.

EXAMPLE:

see example.m

IMPORTANT FUNCTIONS:

Main function for causal discovery:

function [W_hat, thetaW_hat,thetaE_hat, W_save,thetaW_save,thetaE_save,Q_save] = SAEM_ins(Data,thetaW0,thetaE0,np,Mask)

• INPUT:

  • Data: data from each subject are saved in a cell
  • thetaW0: initial values of the parameters related to W (W = I-B and B is the causal adjacency matrix), see equation (7) in the paper
  • thetaE0: initial values of the parameters related to E
  • np: number of particles that need to be sampled
  • Mask: use it to fix some entries of B to zero, where B = I-W

• OUTPUT:

  • W_hat: estimated W for each individual
  • thetaW_hat: estimated parameters related to W
  • thetaE_hat: estimated parameters related to E
  • W_save: sampled W's in each iteration
  • thetaW_save: estimated theta_W in each iteration
  • thetaE_save: estimated theta_E in each iteration
  • Q_save: estimated Q value in each iteration

Main function for mechanism-based clustering:

function Pz = clustering_ins(Data,thetaW,thetaE,Mask,nZ)

• INPUT:

  • Data: data from each subject are saved in a cell
  • thetaW: related parameters of W
  • thetaE: related parameters of E
  • Mask: use it to fix some entries of B to zero, where B = I-W
  • nZ: number of groups

• OUTPUT:

  • Pz: Pz(i,j) the posterior probability that individual i is in group j

Citation:

B. Huang, K. Zhang, P. Xie, M. Gong, E. Xing, C. Glymour. Specific and Shared Causal Relation Modeling and Mechanism-based Clustering. NeurIPS'19

If there are any questons, please send emails to biweih@andrew.cmu.edu