Skip to content

A method which takes advantage of causal features for classification

Notifications You must be signed in to change notification settings

xiaowei-ui/Causally-Regularized-Learning

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

3 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Causally-Regularized-Learning

This repo contains the core code for method described in Zheyan Shen, Peng Cui, Kun Kuang, Bo Li, and Peixuan Chen. 2018. Causally Regularized Learning with Agnostic Data Selection Bias. In 2018 ACM Mul- timedia Conference (MM ’18)

Here is a simple demo showing how it works in Matlab:

% Generate predictor X and outcome Y (binary)
X = 2*round(rand(1000, 20))-1; % 1000 samples and 20 features
beta_true = ones(20, 1);
Y = double(sigmoid(X*beta_true)>=0.5);
lambda0 = 1; %Logistic loss
lambda1 = 0.1; %Balancing loss
lambda2 = 1; %L_2 norm of sample weight
lambda3 = 0; %L_2 norm of beta
lambda4 = 0.001; %L_1 norm of bata
lambda5 = 1; %Normalization of sample weight
MAXITER = 1000;
ABSTOL = 1e-3;
W_init = rand(1000, 1);
beta_init = 0.5*ones(20, 1);
[W, beta, J_loss] = mainFunc(X, Y, ...
        lambda0, lambda1, lambda2, lambda3, lambda4, lambda5,...
        MAXITER, ABSTOL, W_init, beta_init);

About

A method which takes advantage of causal features for classification

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • MATLAB 100.0%