Skip to content

Latest commit

 

History

History
32 lines (22 loc) · 1.04 KB

README.md

File metadata and controls

32 lines (22 loc) · 1.04 KB

cs6362-project

(manuscript available upon request)

Final project for cs6362: advanced machine learning

Motivating question: How do Variational Causal Networks (VCNs) approximation of the posterior over Direct Acyclic Graphs (DAGs) in causal structure learning compare to the boostrapped estimation via Direct LiNGAM in high dimensional settings?

VCN adapted from: https://github.com/yannadani/vcn_pytorch (instructions located under models/vcn_adapted)

Boostrapped DirectLiNGAM adapted from: https://github.com/cdt15/lingam

Data (and verified ground truth source):

  1. https://pubmed.ncbi.nlm.nih.gov/15845847/
  2. https://arxiv.org/abs/1805.03108
Line # Intervention
0-852 cd3_cd28
853-1754 icam2
1755-2665 aktinhib
2666-3388 g0076
3389-4198 psitect
4199-4997 u0126
4998-5845 ly
5846-6758 pma
6759-7465 b2camp

Future work: ELBO gradient --> score function estimator variance reduction for more optimal learning within VCNs using control variates.