Online sampling from log-concave distributions

Authors: Holden Lee, Oren Mangoubi, Nisheeth Vishnoi

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

Reproducibility Variable Result LLM Response
Research Type Experimental In simulations, our algorithm achieves accuracy comparable to an algorithm specialized to logistic regression.
Researcher Affiliation Academia Holden Lee Duke University Oren Mangoubi Worcester Polytechnic Institute Nisheeth K. Vishnoi Yale University
Pseudocode Yes Algorithm 1 SAGA-LD ... Algorithm 2 Online SAGA-LD
Open Source Code No The paper does not contain any explicit statements or links indicating that source code for the described methodology is publicly available.
Open Datasets No The data is generated as follows. First, N(0, Id), b N(0, 1) are randomly generated. For each 1 t T, a feature vector xt 2 Rd and output yt 2 {0, 1} are generated by xt,i Bernoulli yt Bernoulli(φ( >xt + b)), (3) where the sparsity is s = 5 in our simulations, and φ(x) = 1 1+e x is the logistic function. We chose xt 2 {0, 1}d because in applications, features are often indicators. (The paper describes a synthetic dataset generation process but does not make the dataset itself publicly available or refer to a standard public dataset.)
Dataset Splits No The paper describes how samples were collected and simulations replicated (
Hardware Specification Yes The experiments were run on Fujitsu CX2570 M2 servers with dual, 14-core 2.4GHz Intel Xeon E5 2680 v4 processors with 384GB RAM running the Springdale distribution of Linux.
Software Dependencies No The paper mentions 'running the Springdale distribution of Linux' but does not specify any programming languages, libraries, or other software dependencies with version numbers used for the experiments.
Experiment Setup Yes The step size at epoch t is 0.1 1+0.5t for MALA, 0.01 1+0.5t for SGLD, and 0.05 1+0.5t for online SAGA-LD. A smaller step size must be used with SGLD because of the increased variance. For MALA, a larger step size can be used because the Metropolis-Hastings acceptance step ensures the stationary distribution is correct. The batch size for SGLD and online SAGA-LD is 64.