Self-Adversarially Learned Bayesian Sampling

Authors: Yang Zhao, Jianyi Zhang, Changyou Chen5893-5900

AAAI 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments on both synthetic and real datasets verify advantages of our framework, outperforming related methods in terms of both sampling efficiency and sample quality.
Researcher Affiliation Academia Yang Zhao State University of New York at Buffalo yzhao63@buffalo.edu Jianyi Zhang Fudan University 15300180019@fudan.edu.cn Changyou Chen State University of New York at Buffalo cchangyou@gmail.com
Pseudocode Yes Algorithm 1: SAL-MC training and sampling
Open Source Code No The paper states "detailed algorithm given in the Supplementary Material (SM) on our homepage" but does not provide a concrete link to source code for the described methodology.
Open Datasets Yes Experiments on both synthetic and real datasets... for image synthesis on MNIST and Celeb A datasets... We further compare SAL-MC with A-NICE-MC on several BLR tasks... Three datasets, Heart (532-13), Australian (690-14) and German (1000-24), are used...
Dataset Splits No The paper states "The models are trained on a random 80% of the datasets and tested on the remaining 20% in each run" but does not explicitly mention a separate validation split.
Hardware Specification No The paper does not provide specific hardware details (e.g., GPU/CPU models, memory) used for running its experiments.
Software Dependencies No The paper does not provide specific ancillary software details, such as library names with version numbers, needed to replicate the experiment.
Experiment Setup Yes As suggested by (Welling and Teh 2011), a polynomially-decayed step size ϵt = a/(t + 1)0.55 is used in SGLD for a fair comparison... The mini-batch size for training is 64; and the injected noise ξ is drawn from N(0, I) for all tasks.