Hello Universe !

May 13, 2019

Introduction :

Hello, I’m Ronit Jorvekar, computer engineering undergraduate from Pune Institute of Technology, Pune India. I’m passionate about strategically drawing inferences from data, programming and software engineering design. For my GSoC project this year, I would be working on graph neural networks to determine the social acceptability score of a robot in a particular environment.

Project:

Learning socially acceptable behavior using machine learning techniques on graph data.

About graph data :

Recent interest in the field of machine learning is focused on Euclidean data. Graph data is different as it is non Euclidean. It consists of nodes and edges between the nodes. Certain machine learning techniques are used to generate embeddings for the nodes and hence represent them in euclidean form.

Scenario represented in graph :

Graph data is better able to extract the semantics from environment. It represents node features as well as relations. With the help of inherent features of the node and the Relational data , we can generate inferences. The score is determined by taking into consideration the social conventions.

This project aims at applying recent machine learning techniques to create model which predicts the social acceptability score of the robot given a scenario represented in graph. It would be able to predict the score in the range of 0-100. The model I would be focusing on is RGCN model. The ability to predict the social acceptability score would be added as a new component . Other components would be able to access the score generated by this component through an interface.

Ronit Jorvekar