Sample-Efficient Geometry Reconstruction from Euclidean Distances using Non-Convex Optimization

Authors: Ipsita Ghosh, Abiy Tasissa, Christian Kümmerle

NeurIPS 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Furthermore, we assess data efficiency, scalability and generalizability of different reconstruction algorithms through numerical experiments with simulated data as well as real-world data, demonstrating the proposed algorithm s ability to identify the underlying geometry from fewer distance samples compared to the state-of-the-art.
Researcher Affiliation Academia Ipsita Ghosh Department of Computer Science University of North Carolina at Charlotte ighosh2@charlotte.edu Abiy Tasissa Department of Mathematics Tufts University Abiy.Tasissa@tufts.edu Christian Kümmerle Department of Computer Science University of North Carolina at Charlotte kuemmerle@charlotte.edu
Pseudocode Yes Algorithm 1 Matrix IRLS for Euclidean Distance Geometry Problems
Open Source Code Yes The Matlab code can be found at github_EDG-IRLS
Open Datasets Yes For our experiments, we determine the structures of a protein molecule (1BPM) from the Protein Data Bank [HMBFGG+00]
Dataset Splits No No explicit mention of training, validation, or test dataset splits (e.g., percentages or sample counts) was found. The paper describes random sampling of distance pairs for input, not traditional dataset splits for train/validation/test.
Hardware Specification Yes The experiements were performed on a single node of a compute cluster equipped with dual 24-core Intel Xeon Gold 6248R CPUs, utilizing 32 parallel tasks.
Software Dependencies No The paper mentions using 'Matlab code' and adapts implementations for 'Scaled SGD', 'ALM', and 'Rie EDG', but does not provide specific version numbers for these software dependencies or libraries.
Experiment Setup Yes The input parameter for Matrix IRLS includes the number of outer IRLS iterations N 0, the number of inner iterations N 0 inner, the tolerance, which is the stopping criteria for the algorithm tol0 inner. For large datasets like the UScities data and Protein data N = 400, although the algorithm converges within 120 iterations. We run the experiments with N 0 inner = 2000 and tolinner = 10 10. In a smaller setup of the synthetic data we perform the experiments with n = 500 points with same parameters. although convergence of Matrix IRLS is observed much faster. To put more emphasis on the per iteration error, we study the experiment on per-iterate analysis with more iterations.