Developmental approaches to hybrid-machine-learning

Improve efficiency for hybrid-machine learning using developmental-learning strategies.
Description of the Project: 

The aim of this project is to improve the efficiency and understanding of current Machine Learning (ML) techniques by taking developmental approaches into consideration.

This will be accomplished through a serious of child-parents and human-robot interaction developing an appropriated data set, to developing a learning algorithm that thrives from social rules and semantic constrains of the learning input. Through the use of state-of-the-art technology, e.g. the iCub robot and the DGX learning cluster, this project will have the benefit of developing novel ideas of how to overcome data limitations and get the chances to explore possible future algorithms.

Furthermore, in the interface between development physiology and robotics, this interdisciplinary project is envisaged to go beyond the current big-data ML approaches into a more efficient continuous learning approach.

Resources required: 
DGX, iCub robot
Project number: 
First Supervisor: 
Heriot-Watt University
First supervisor university: 
Heriot-Watt University
Essential skills and knowledge: 
C++, Python, deeplearning, HRI, ROS
Desirable skills and knowledge: 
Yarp, iCub