Robust Learning from Noisy Side-information by Semidefinite Programming

Authors: En-Liang Hu, Quanming Yao

IJCAI 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Finally, empirical study shows that the new objective armed with proposed algorithm outperforms state-of-the-arts in terms of both speed and accuracy.
Researcher Affiliation Collaboration En-Liang Hu1 , Quanming Yao2,3 1Department of Mathematic, Yunnan Normal University 24Paradigm Inc 3Department of Computer Science and Engineering, Hong Kong University of Science and Technology ynel.hu@gmail.com, yaoquanming@4paradigm.com
Pseudocode Yes Algorithm 1 RSDP: Robust semi-definite programming by majorization-minimization. 1: Initialization: X1 = 0. 2: for k = 1, . . . , K do 3: Xk =arg min X H( X, Xk) via ADMM or APG; 4: update Xk+1 = Xk + Xk; 5: end for 6: return XK+1.
Open Source Code No Availability of codes and data sets are in Appendix.B. The provided text does not contain Appendix B or a direct link/statement about open-sourcing the code described in the paper.
Open Datasets Yes Experiments are performed on the adult data sets that has been commonly used as benchmark data about NPKL learning [Zhuang et al., 2011].
Dataset Splits Yes We randomly sample 20% pairs from T for training, 20% for validation, and the rest for testing.
Hardware Specification Yes Finally, all algorithms are implemented in Matlab run on a PC with a 3.07GHz CPU and 24GB RAM.
Software Dependencies No The paper mentions 'Matlab' and various algorithms/packages (e.g., 'FW', 'L-BFGS', 'nm APG', 'SADMM', 'SDPNAL', 'SDPLR') but does not specify version numbers for any of them.
Experiment Setup Yes All algorithm is stopped when the relative change of objective values in successive iterations is smaller than 10 5 or when the number of iterations reaches 2000. As for the rank r of initial solution X, in Sections 4.1 and 4.2 we follow [Burer and Monteiro, 2003] and set its value to be the largest r satisfying r(r + 1) m, where m is the total number of observed data (i.e, m is the number of must-link and cannotlink pairs in Section 4.1, the number of given neighbor pairs in Section 4.2 respectively). In Section 4.3 we set r = 10. The tradeoff parameter γ is set to 0.01 as a default. We set γ = 10 to obtain sparse solution.