Handling concept drift in time series datastream of a robot

Project Overview

This project is about evaluating drift detection algorithm
in the context of detecting new terrain of a hexapod robot.
The dataset consists of six types of terrains : Black Flat, Blacks Rough, Flat, Wooden Cube, Black Rough, Grass Rough

Content of the repository

ADWIN

ADWIN is a drift detection algorithm with the based principle of shrinking the window whever a change occurs. For more
information, check 'Learning from time-changing data with Adaptive Window' from Albert Bifet. The code is inspired from the C++ version written by the author and available at : https://github.com/abifet/adwin/tree/master/C%2B%2B

Page-Hinkey Test

P-H test is a drift detection algorithm design for gradual changes.
The code has been inspired from : "Knowledge Discovery from Data Streams."

Kolmogorov

We have done a light adaptation of the K-s test to the context of Drift detection.
For a more detailed description of this work. See the master's 1 thesis link to it Presentation slides