Distributed Estimation, Information Loss and Exponential Families
Authors: Qiang Liu, Alex Ihler
NeurIPS 2014 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | 5 Experiments and Practical Issues. We present numerical experiments to demonstrate the correctness of our theoretical analysis. More importantly, we also study empirical properties of the linear and KL combination methods that are not enlightened by the asymptotic analysis. We find that the linear average tends to degrade significantly when its local models (ˆθk) are not already close, for example due to small sample sizes, heterogenous data partitions, or non-convex likelihoods... We experiment on the MNIST dataset and the Year Prediction MSD dataset in the UCI repository... |
| Researcher Affiliation | Academia | Qiang Liu Alexander Ihler Department of Computer Science, University of California, Irvine qliu1@uci.edu ihler@ics.uci.edu |
| Pseudocode | No | The paper does not contain structured pseudocode or algorithm blocks. |
| Open Source Code | No | The paper does not provide concrete access to source code (specific repository link, explicit code release statement, or code in supplementary materials) for the methodology described in this paper. |
| Open Datasets | Yes | We experiment on the MNIST dataset and the Year Prediction MSD dataset in the UCI repository, where the training data is partitioned into 10 sub-groups randomly and evenly. |
| Dataset Splits | Yes | We experiment on the MNIST dataset and the Year Prediction MSD dataset in the UCI repository, where the training data is partitioned into 10 sub-groups randomly and evenly. In both cases, we use the original training/test split; we use the full testing set, and vary the number of training examples n by randomly sub-sampling from the full training set (averaging over 100 trials). ... For MNIST, we also consider a severely heterogenous data partition by splitting the images into 10 groups according to their digit labels. |
| Hardware Specification | No | The paper does not provide specific hardware details (exact GPU/CPU models, processor types with speeds, memory amounts, or detailed computer specifications) used for running its experiments. |
| Software Dependencies | No | The paper does not provide specific ancillary software details (e.g., library or solver names with version numbers) needed to replicate the experiment. |
| Experiment Setup | Yes | The number of mixture components is fixed to be 10. ... We take the first 100 principal components when using MNIST. |