Message Passing Least Squares Framework and its Application to Rotation Synchronization
Authors: Yunpeng Shi, Gilad Lerman
ICML 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We demonstrate the superior performance of our algorithm over state-of-the-art methods for rotation synchronization using both synthetic and real data. |
| Researcher Affiliation | Academia | 1School of Mathematics, University of Minnesota, Minneapolis, MN, USA. Correspondence to: Yunpeng Shi<shixx517@umn.edu>, Gilad Lerman<lerman@umn.edu>. |
| Pseudocode | Yes | Algorithm 1 CEMP (Lerman & Shi, 2019) ; Algorithm 2 Message Passing Least Squares (MPLS) ; Algorithm 3 MPLS-SO(3) |
| Open Source Code | No | The paper mentions using implementations by other researchers ('We use their implementation by Chatterjee & Govindu (2018)') but does not provide access to the authors' own source code for the described methodology. |
| Open Datasets | Yes | We compare the performance of the different algorithms on the Photo Tourism datasets (Wilson & Snavely, 2014). |
| Dataset Splits | No | The paper does not explicitly provide specific training/test/validation dataset splits (e.g., percentages, sample counts, or references to predefined splits) needed for reproduction. |
| Hardware Specification | Yes | All computational tasks were implemented on a machine with 2.5GHz Intel i5 quad core processors and 8GB memory. |
| Software Dependencies | No | The paper mentions using existing implementations by other authors (e.g., Chatterjee & Govindu), but it does not specify software dependencies with version numbers for its own implementation or experiments. |
| Experiment Setup | Yes | We use the following default parameters for Algorithm 1: |Cij|= 50 for ij E; T =5; βt=2t and t=0,...,5. If an edge is not contained in any 3-cycle, we set its corruption level as 1. For MPLSSO(3), which we refer to in this section as MPLS, we use the above parameters of Algorithm 1 and the following ones for t 1: αt=1/(t+1) and τt=inf x n ˆPt(x)>max{1 0.05t,0.8} o . ... F(x) for MPLS is chosen as x 3/2 and it corresponds to ρ(x)= x. ... We use the convergence criterion P i [n] Ωi,t F/( 2n)<0.001 of Chatterjee & Govindu (2018) for all the above algorithms. |