Skip to content

Latest commit

 

History

History
22 lines (14 loc) · 1.36 KB

README.md

File metadata and controls

22 lines (14 loc) · 1.36 KB
Raw

MATLAB implementaion of Federated Over-the-Air PCA and Subspace Learning from Incomplete data

If you use the codes please cite the following paper

[1] "Federated Over-the-Air Subspace Learning from Incomplete Data", Praneeth Narayanamurthy, Namrata Vaswani, Aditya Ramamoorthy, https://arxiv.org/abs/2002.12873

List of files:

Fully observed data, static subspace setting (FedPM)

  1. taubatch_test.m: this file considers the case when we are allowed to vary $\tau$ -- the number of iterations after which we normalize the power method output.
  2. sigma_c_tes.m: this file has the codes for varying the channel noise seen at each iteration.
  3. ratio_test.m: this file studies the variation in the eigen ratio.

(dynamic) Subspace Tracking with Missing Entries (FedSTMiss)

  1. NORST_fed.m: this script contains the function to implement Algorithm 3. This tracks time-varying subspaces, deals with noise, and provides a "federated, over the air implementation".
  2. st_miss_fed.m: this is the wrapper to generate data for st-miss problem, and implement FedSTMiss.
  3. simple_evd, ccgls, calc_subspace_error, cgls, phifun.m: helper functions used inside NORST_fed.m
  4. PROPACK: linear algebra toolbox (downloaded from https://sun.stanford.edu/~rmunk/PROPACK/)

##need to add real data experiments, should be straightforward, but will have to figure out what to compare with.