A Bayesian Inference Benchmark Suite
If you find BayesSuite useful, please cite our paper:
Yu Wang, Yuhao Zhu, Glenn G. Ko, Brandon Reagen, Gu-Yeon Wei, and David Brooks. “Demystifying Bayesian Inference Workloads.” International Symposium on Performance Analysis of Systems and Software (ISPASS), 2019.
The paper profiles the computational characteristics of BayesSuite using single-core and multi-core CPU. It also has a last level cache miss prediction model based on the model and the datasets.
|12cities||Poisson Regression||Does lowering speed limits save pedestrian lives?|
|ad||Logistic Regression||Advertising attribution in the movie industry|
|ode||Friberg-Karlsson Semi-Mechanistic||Solving ordinary differential equations of non-linear systems|
|memory||Hierarchical Bayesian||Modeling memory retrieval in sentence comprehension|
|votes||Hierarchical Gaussian Processes||Forecasting presidential votes|
|tickets||Logistic Regression||Do police officers alter the ticket writing to match departmental targets?|
|disease||Logistic Regression||Measuring the continually worsening progression of Alzheimer’s disease|
|racial||Hierarchical Bayesian||Testing for racial bias in vehicle searches by police|
|butterfly||Hierarchical Bayesian||Estimating butterfly species richness and accumulation|
|survival||Cormack-Jolly-Seber||Estimating animal survival probabilities|
- After git clone this repo, please run
git submodule init git submodule update
- To create BayesSuite and generate run scripts
cd scripts bash start.sh
- To install R packages
If getting publickey issues when running
git submodule update, you need to generate a new SSH key and add it to the ssh-agent and add the new SSH key to your GitHub account.
All of the example models are copyrighted by their author(s). The scripts in this repository are under Apache License 2.0.
Yu (Emma) Wang 3/14/2019