Welcome to the repository of GraphDA! This repository is only for reproducing all experimental results shown in our KDD paper. To install it via pip, please try sparse-learn. More details of GraphDA can be found in: "Zhou, Baojian, Feng Chen, and Yiming Ying. "Dual Averaging Method for Online Graph-structured Sparsity." arXiv preprint arXiv:1905.10714 (2019).".
Our code is based on Openblas-0.3.1, which we already copied into our repository. Suppose you are using GNU/Linux based system or Mac, you can first goto OpenBLAS-0.3.1 folder and then make install it via the following command:
>>> cd OpenBLAS-0.3.1
>>> make && make install PREFIX=../lib
To download the datasets and results, please goto: datasets and results.
The lib folder under graph-da and corresponding libraries will be generated.
To generate Figure 1, run the following command:
>>> python exp_logit_benchmark.py show_figure_1
To generate Figure 2, run the following command:
>>> python exp_logit_benchmark.py show_figure_2
To generate Figure 3-7, run the following command:
>>> python exp_logit_benchmark.py show_figure_3-7
To generate Figure 8, run the following command:
>>> python exp_linear_mnist.py show_figure_8
To generate Figure 9-10, run the following command:
>>> python exp_logit_kegg.py show_figure_9-10
To generate Figure 11, run the following command:
>>> python exp_linear_mnist.py show_figure_11
To generate Figure 12, run the following command:
>>> python exp_logit_kegg.py show_figure_12
To generate Figure 12, run the following command:
>>> python exp_logit_benchmark.py show_4_tables
To reproduce those results, you need to run the following commands:
>>> python exp_logit_benchmark.py run_{fix_tr_mu,diff_tr,diff_mu,diff_s}
>>> python exp_logit_kegg.py {test_graphda,test_baselines}
>>> python exp_linear_mnist.py run_test