https://github.com/sharanry/uncertaintyquantification

Code for Connection between Uncertainty Quantification and Gaussian Prior Parameters
https://github.com/sharanry/uncertaintyquantification

bayesian-neural-networks uncertainty-quantification

Last synced: about 2 months ago
JSON representation

Code for Connection between Uncertainty Quantification and Gaussian Prior Parameters

• Host: GitHub
• URL: https://github.com/sharanry/uncertaintyquantification
• Owner: sharanry
• License: mit
• Created: 2019-09-09T13:43:23.000Z (over 5 years ago)
• Default Branch: master
• Last Pushed: 2023-07-06T21:45:32.000Z (almost 2 years ago)
• Last Synced: 2025-01-22T02:47:00.680Z (4 months ago)
• Topics: bayesian-neural-networks, uncertainty-quantification
• Language: Python
• Size: 768 KB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 2
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# Connection between Uncertainty Quantification and Gaussian Prior Parameters

Supplementary code for the paper on *Connection between Uncertainty Quantification and Gaussian Prior Parameters*.

### Prerequisites

 - [Python3](https://www.python.org/download/releases/3.0/)

 - [virtualenv](https://packaging.python.org/guides/installing-using-pip-and-virtual-environments/)

### Instructions

 - Clone the repository.

 ```bash

git clone https://github.com/sharanry/UncertaintyQuantification

cd UncertaintyQuantification

 ```

 - Make a new `virtualenv` environment and install dependencies from `requirements.txt`

 ```bash

python3 -m virtualenv env

source ./env/bin/activate

pip install -r requirements.txt

 ```

 - Run the two scripts to reproduce results.

 ```bash

 python asymptotic_analysis.py

 python sigma_analysis.py

 ``` 

 - The generated plots will be available in `./results`.

 

### Results

#### Sigma analysis of discontinuous regression using Bayesian neural networks

![](./images/sigma_ci.png)

*Fig 1(a): 95% confidence interval of the models trained over varying prior σ. Red crosses denote the data samples, blue dots represent the mean predictions and the blue fill expresses the 95% CI.*

![](./images/sigma_ess.png)

*Fig 1(b): Effective sample size of weights posterior samples at each layer in logarithmic scale.*

#### Asymptotic analysis of discontinuous regression using Bayesian neural networks

![](./images/asymp_ci.png)

*Fig 2(a): 95

% confidence interval of the models trained over varying data sample sizes. Red crosses denote the data samples with prior σ = 1.0, blue dots represent the mean predictions and the blue fill expresses the 95% CI.*

![](./images/asymp_ess.png)

*Fig 2(b): Effective sample size of weights posterior samples at each layer in logarithmic scale.*