On the Anatomy of MCMC-Based Maximum Likelihood Learning of Energy-Based Models
Authors: Erik Nijkamp, Mitch Hill, Tian Han, Song-Chun Zhu, Ying Nian Wu5272-5280
AAAI 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Our results are shown in Figure 7, Figure 8 (left), and Figure 2 (top). Our recent companion work (Nijkamp et al. 2019) thoroughly explores the capabilities of noise-initialized non-convergent ML. |
| Researcher Affiliation | Academia | Erik Nijkamp,* Mitch Hill,* Tian Han, Song-Chun Zhu, Ying Nian Wu UCLA Department of Statistics 8117 Math Sciences Bldg. Los Angeles, CA 90095-1554 (*equal contributions) |
| Pseudocode | Yes | Algorithm 1: ML Learning |
| Open Source Code | No | The paper mentions a companion work (Nijkamp et al. 2019) that explores capabilities, but it does not explicitly state that the source code for the described methodology in *this* paper is released or provide a link to a repository. |
| Open Datasets | Yes | From left to right: MNIST, Oxford Flowers 102, Celeb A, CIFAR-10. |
| Dataset Splits | No | The paper mentions using a 'training set' and 'batch images' in Algorithm 1, but it does not specify explicit training/validation/test splits, percentages, or sample counts for dataset partitioning. |
| Hardware Specification | No | The paper mentions funding from 'Extreme Science and Engineering Discovery Environment (XSEDE) grant ASC170063', implying computational resources were used, but it does not provide specific hardware details like CPU/GPU models, processor types, or memory. |
| Software Dependencies | No | The paper mentions optimizers ('SGD or Adam (Kingma and Ba 2015)') and discusses convolutional neural networks, but it does not specify any software names with version numbers (e.g., Python 3.x, PyTorch 1.x, TensorFlow 2.x) to enable reproduction of the environment. |
| Experiment Setup | Yes | For non-convergent training we find the tuning of noise and stepsize have little effect on training stability. We use ε = 1 and τ = 0. |