Alleviating Label Switching with Optimal Transport
Authors: Pierre Monteiller, Sebastian Claici, Edward Chien, Farzaneh Mirzazadeh, Justin M. Solomon, Mikhail Yurochkin
NeurIPS 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We give conditions under which there is a meaningful solution to label switching and demonstrate advantages over alternative approaches on simulated and real data. |
| Researcher Affiliation | Collaboration | Pierre Monteiller ENS Ulm pierre.monteiller@ens.fr Sebastian Claici MIT CSAIL & MIT-IBM Watson AI Lab sclaici@mit.edu Edward Chien MIT CSAIL & MIT-IBM Watson AI Lab edchien@mit.edu Farzaneh Mirzazadeh IBM Research & MIT-IBM Watson AI Lab farzaneh@ibm.com Justin Solomon MIT CSAIL & MIT-IBM Watson AI Lab jsolomon@mit.edu Mikhail Yurochkin IBM Research & MIT-IBM Watson AI Lab mikhail.yurochkin@ibm.com |
| Pseudocode | Yes | Algorithm 1 Riemannian Barycenter of Ω. Input: Distribution Ω, exp and log maps on M Output: Estimate of the barycenter of Ω |
| Open Source Code | No | The paper mentions using a third-party tool (Stan) but does not provide open-source code for their own methodology. |
| Open Datasets | No | The paper describes experiments on 'simulated and real data' but does not specify access information (e.g., links, citations, or repository names) for publicly available datasets used for training or general data analysis. |
| Dataset Splits | No | The paper discusses data generation and sampling processes (e.g., 'four chains discarding 500 burn-in samples'), but does not describe conventional train/validation/test dataset splits for reproducibility. |
| Hardware Specification | No | The paper does not provide any specific hardware details such as GPU models, CPU types, or memory specifications used for running the experiments. |
| Software Dependencies | No | The paper mentions using 'Stan (Carpenter et al., 2017)' as an HMC sampler but does not provide specific version numbers for Stan or other software dependencies, which are necessary for reproducibility. |
| Experiment Setup | Yes | We implement HMC sampler using Stan (Carpenter et al., 2017), with four chains discarding 500 burn-in samples and keeping 500 per chain. |