Optimal Estimator for Unlabeled Linear Regression

Authors: Hang Zhang, Ping Li

ICML 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Numerical experiments are also provided to corroborate the theoretical claims. 5. Simulations This section presents the numerical results. Since our estimator cannot guarantee the correct permutation matrix Π under the single observation model, our simulations focus on the multiple observations model, i.e., m > 1.
Researcher Affiliation Industry Hang Zhang, Ping Li Cognitive Computing Lab Baidu Research 10900 NE 8th ST. Bellevue, WA 98004, USA {zhanghanghitomi, pingli98}@gmail.com
Pseudocode Yes Algorithm 1 The one-step estimator proposed in this paper. Input: observation Y and sensing matrix X. Output: pair ( !Π, !B), which is written as !Π = argmaxΠ Pn !B = (X) !Π where X = (X X) 1X is the pseudo-inverse of X and Pn is the set of all possible permutation matrices.
Open Source Code No The paper does not provide any statement or link indicating the availability of open-source code for the described methodology.
Open Datasets No The paper describes generating synthetic data for simulations: 'We closely follow the experiment setting in Zhang et al. (2019b). We set the i-th column B :,i (1 i min(m, p)) to be the i-th canonical basis, which has 1 on the i-th entry and 0 elsewhere.' It does not refer to a publicly available dataset for training.
Dataset Splits No The paper describes simulation experiments but does not provide explicit train/validation/test dataset splits or cross-validation details.
Hardware Specification No The paper does not provide any specific hardware details (e.g., GPU/CPU models, memory) used for running the experiments.
Software Dependencies No The paper does not specify any software dependencies with version numbers.
Experiment Setup Yes For each n, we choose p {0.1n, 0.2n} and h {n/10, n/4}. That is, when n = 500, we have p {50, 100} and h {50, 125}; and when n = 1000, we have p {100, 200} and h {100, 250}. For each chosen set of parameters (n, p, m, h) and SNR value, we simulate the data 1000 times and report the success rate of exact recovery of Π using our proposed estimator in Algorithm 1.