Convergence Guarantees for a Class of Non-convex and Non-smooth Optimization Problems

Authors: Koulik Khamaru, Martin J. Wainwright

JMLR 2019 | Venue PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We illustrate our methods and theory via applications to the problems of best subset selection, robust estimation, mixture density estimation, and shape-from-shading reconstruction. ... Some reconstruction experiments: In order to illustrate the behavior of our method for this problem, we considered two synthetic images for simulated experiments. ...Figure 2 shows the performances of the prox-type method and CCCP for synthetic data simulated as above, with problem parameters (n, p) = (190, 300) and (n, p) = (380, 600) and different choices of sparsity s.
Researcher Affiliation Collaboration Koulik Khamaru EMAIL Martin J. Wainwright , , EMAIL Department of Statistics Department of Electrical Engineering & Computer Sciences University of California Berkeley Berkeley, CA 94720-1776, USA Voleon Group Berkeley, CA
Pseudocode Yes Algorithm 1 Subgradient-type method Algorithm 2 Proximal-type algorithm Algorithm 3 Frank-Wolfe type method Algorithm 4 Gradient descent with backtracking
Open Source Code No The paper does not provide any explicit statements about releasing code or links to source code repositories.
Open Datasets Yes The first one is a 256 256 image of Mozart (Zhang et al., 1999), and the second one is a 128 128 image of Vase.
Dataset Splits No The paper describes the generation of synthetic data but does not specify how this data was split into training, validation, or test sets for experiments.
Hardware Specification No The paper mentions runtimes for experiments ("The runtime for Mozart-example was 87 seconds, whereas the runtime for Vase-example was 39 seconds") but does not provide specific details about the hardware (e.g., CPU, GPU models, memory) used.
Software Dependencies No The paper describes various algorithms and their theoretical properties. While it discusses implementation aspects, it does not specify any software libraries, frameworks, or their version numbers used for the experiments.
Experiment Setup Yes For both the algorithms, the tolerance level η was set to η = 10 8, whereas the maximum number of iterations was 1000. In all our simulations we used same initializations for both the algorithms.