Mixed Regression via Approximate Message Passing

Authors: Nelvin Tan, Ramji Venkataramanan

JMLR 2023 | Venue PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental The theoretical results are validated by numerical simulations for mixed linear regression, max-affine regression, and mixture-of-experts. For max-affine regression, we propose an algorithm that combines AMP with expectation-maximization to estimate the intercepts of the model along with the signals. The numerical results show that AMP significantly outperforms other estimators for mixed linear regression and max-affine regression in most parameter regimes.
Researcher Affiliation Academia Nelvin Tan EMAIL Department of Engineering, University of Cambridge Cambridge, CB2 1PZ, United Kingdom Ramji Venkataramanan EMAIL Department of Engineering, University of Cambridge Cambridge, CB2 1PZ, United Kingdom
Pseudocode Yes Algorithm 1 Expectation-maximization approximate message passing (EM-AMP)
Open Source Code No The paper does not contain any explicit statements or links indicating that the source code for the methodology described is publicly available or released. It mentions a preliminary version published in AISTATS 2023 but does not provide code access.
Open Datasets No The paper describes generating synthetic data for its simulations (e.g., "Xi i.i.d. N(0, Ip/n)", "ci i.i.d. Bernoulli(α)"). It does not refer to or provide access information for any existing public datasets.
Dataset Splits No The paper describes generating synthetic data for its numerical simulations and varies parameters like 'p' (signal dimension) and 'n' (number of observations) to control the sampling ratio 'δ = n/p'. It does not discuss conventional training/validation/test splits for specific datasets.
Hardware Specification No The paper does not provide any specific details about the hardware used for running the experiments, such as GPU/CPU models, processors, or memory specifications.
Software Dependencies No The paper does not explicitly mention any specific software dependencies or their version numbers that would be required to reproduce the experiments.
Experiment Setup Yes In Figures 1, 2, and 3, we set the Bernoulli parameter α = 0.7 and choose the two signals to be jointly Gaussian, with their entries generated as (β(1) j , β(2) j ) i.i.d. N( 0 0 , j [p]). The signal dimension p = 500 and vary the value of n in our experiments. For the soft-thresholding denoiser fk... The tuning parameter ζ set to 1.1402. ...we execute EM-AMP with mmax = 5 and kmax = 5.