MichalTesnar / mystery

Uncertainty in Active Incremental Online Learning

Welcome to the repository of my thesis! I hope you will be enlightened by the awesome things that you will learn here!

Project Content

In this project, we investigate the use of uncertainty quantification in the online incremental learning setting in two tasks. One is an artificial dataset generated by a sinus curve. The other

How to setup?

Get Python 3.

I will leave that up to you...

Clone the repo.

git clone https://github.com/MichalTesnar/mystery

Get Keras Uncertainty, the library for uncertainty quantification in Keras.

pip install --user git+https://github.com/mvaldenegro/keras-uncertainty.git

Install dependencies.

pip install -r dependencies.txt

Configure the project

Altering the experiment_specification variable in main.py to change hyperparameters. Change the dataset object to be either SinusiodToyExample or DagonAUVDataset, depending on the dataset of choice.

RUN!

python3 main.py

Your resulting plots will be saved in /results.