Streaming, Distributed Variational Inference for Bayesian Nonparametrics
Authors: Trevor Campbell, Julian Straub, John W. Fisher III, Jonathan P. How
NeurIPS 2015 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | In this section, the proposed inference framework is evaluated on the DP Gaussian mixture with a normal-inverse-Wishart (NIW) prior. We compare the streaming, distributed procedure coupled with standard variational inference [24] (SDA-DP) to five state-of-the-art inference algorithms: memoized online variational inference (mo VB) [13], stochastic online variational inference (SVI) [9] with learning rate (t+10) 1/2 , sequential variational approximation (SVA) [7] with cluster creation threshold 10−1 and prune/merge threshold 10−3, subcluster splits MCMC (SC) [14], and batch variational inference (Batch) [24]. ... Figure 3 shows the results from the experiment over 30 trials... |
| Researcher Affiliation | Academia | Trevor Campbell1 Julian Straub2 John W. Fisher III2 Jonathan P. How1 1LIDS, 2CSAIL, MIT {tdjc@ , jstraub@csail. , fisher@csail. , jhow@}mit.edu |
| Pseudocode | No | The paper describes the algorithm steps in narrative and with diagrams (Figure 1), but does not provide a formally structured pseudocode block or a section explicitly labeled 'Algorithm'. |
| Open Source Code | No | The paper states 'For the experiments in this work, we used the implementation at github.com/hrldcpr/hungarian.' which refers to a third-party library used, not the open-sourcing of the authors' own methodology. |
| Open Datasets | Yes | MNIST Digits [25]: This dataset consisted of 70,000 28 × 28 images of hand-written digits, with 10,000 held out for testing. ... SUN Images [26]: This dataset consisted of 108,755 images from 397 scene categories, with 8,755 held out for testing. ... [25] Yann Le Cun, Corinna Cortes, and Christopher J.C. Burges. MNIST database of handwritten digits. Online: yann.lecun.com/exdb/mnist. [26] Jianxiong Xiao, James Hays, Krista A. Ehinger, Aude Oliva, and Antonio Torralba. SUN 397 image database. Online: vision.cs.princeton.edu/projects/2010/SUN. |
| Dataset Splits | No | The paper specifies held-out test sets for MNIST (10,000), SUN (8,755), and Airplane Trajectories (1,000), but it does not explicitly state details for a separate validation set or its split percentage. |
| Hardware Specification | No | The paper mentions 'All experiments were performed on a computer with 24 CPU cores and 12Gi B of RAM,' which describes the general hardware specifications but lacks specific details like CPU model, GPU model, or clock speeds. |
| Software Dependencies | No | The paper mentions using 'github.com/hrldcpr/hungarian' but does not provide specific version numbers for this or any other software dependencies like programming languages or libraries. |
| Experiment Setup | Yes | SDA-DP minibatch inference was truncated to K = 50 components, and all other algorithms were truncated to K = 200 components. ... Stochastic variational inference (SVI) [9] with learning rate (t+10) 1/2 , sequential variational approximation (SVA) [7] with cluster creation threshold 10−1 and prune/merge threshold 10−3 ... minibatches of size 50. ... Data was split into minibatches of size 100... Data was split into minibatches of size 500. |