Skip to content

Jedrek00/Dropout-models-uncertainty

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Dropout-models-uncertainty

Experiments for modeling uncertainty with dropout research.
Project made for Deep Learning classes
Date of completion: 📆 24.01.2024 📆

About The Project

The goal of the project was to create relevant experiments to extend the scientific work Gal, Y and Ghahramani, Z. Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learning. We reproduced the research described in the aforementioned paper, and also checked whether we were able to obtain similar results using other dropout techniques.

We will soon describe the results of the experiments in a corresponding blog post!

Technologies Used

Python MLFlow Matplotlib NumPy PyTorch

Installation

Poetry is used to manage dependencies for this repository. Check if poetry is installed on your machine using command:

poetry --version

If version of the poetry was displayed, you can install all required packages using command:

poetry install

To install poetry on your machine see documentation.

All code should be ran in the env created by the poetry. To ensure that you are running scripts in the env created by poetry run command:

poetry shell

Installing new modules

Instead of using pip install module_name Use:

poetry add module_name

MLFlow

Run MLFlow tool inside virtual environment created by poetry:

mlflow ui

Example results

The chart below illustrates the performance of our algorithm for arbitrarily chosen hyperparameters. This is a perfect example of correctly modeling uncertainty using the dropout technique.

Modelling uncertainty example plot

Our project allows user to generate its own, custom charts for user-selected hyperparameters.

About

No description, website, or topics provided.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Contributors 3

  •  
  •  
  •  

Languages