Avatar

Yordan Hristov

PhD student

University of Edinburgh

Biography

PhD student at the Robust Autonomy and Decisions group, part of the Insitute for Perception, Action and Behaviour at the University of Edinburgh. I am supervised by Dr. Subramanian Ramamoorthy and Prof. Alex Lascarides.

My research interests stem from the field of Human-Robot Interaction, and revolve around achieving semantic alignment between human experts and robotic agents. I work on methods for physical symbol grounding and neural semi-supervised representation learning, utilising task-informed inductive biases.

Interests

  • Human-Robot Interaction
  • Learning from Demonstration
  • Disentanglement Learning
  • (Physical) Symbol Grounding

Education

  • PhD in Robotics and Autonomous Systems, 2021

    University of Edinburgh

  • BSc with Honours in Computer Science, 2016

    University of Edinburgh

Publications

Learning from Demonstration with Weakly Supervised Disentanglement

Robotic manipulation tasks, such as wiping with a soft sponge, require control from multiple rich sensory modalities. Human-robot …

Composing Diverse Policies for Temporally Extended Tasks

Robot control policies for temporally extended and sequenced tasks are often characterized by discontinuous switches between different …

Disentangled Relational Representations for Explaining and Learning from Demonstration

Learning from demonstration is an effective method for human users to instruct desired robot behaviour. However, for most non-trivial …

Hybrid system identification using switching density networks

Behaviour cloning is a commonly used strategy for imitation learning and can be extremely effective in constrained domains. However, in …

Using Causal Analysis to Learn Specifications from Task Demonstrations

Learning models of user behaviour is an important problem that is broadly applicable across many application domains requiring …

Experience

 
 
 
 
 

Research Associate

University of Edinburgh

Aug 2020 – Present Edinburgh, UK
Part of a project on Trustworthy Autonomy. Apply different inductive biases and model structures - Graph NNs, Koopman theory - to the problem of robustly modelling nonlinear dynamics of deformable objects.
 
 
 
 
 

Machine Learning Intern

Amazon Development Center Scotland

Jun 2017 – Oct 2017 Edinburgh, UK
Did named entity disanbiguation utilising high-parameter neural models.
 
 
 
 
 

Software Engineering Intern

Skyscanner

Jun 2016 – Aug 2016 Edinburgh, UK
Developed internal tool for asset management in the Hotels dev team.
 
 
 
 
 

Technology Summer Analyst

Bank of America Merrill Lynch

Jun 2015 – Aug 2015 London, UK
Worked on tools for analysing and reporting the throughput of the software of the Global Markets Technology division

Blog

Disentanglement != Axis Alignment

TL;DR - The bottleneck layer of a VAE might be disentangled, up to a rotation, which might make it look entangled upon the standard …

Contact

  • Informatics Forum, 10 Chrichton St, Edinburgh, Scotland, EH8 9AB