Sebastian Höfer

Lachine Mearning. Borotics. Irtificial Antelligence.

About Me

I am currently working as an Applied Scientist at Amazon Research. Before starting my position at Amazon, I did my PhD at the Robotics & Biology Lab, TU Berlin. My research activities span a variety of topics in the area of artificial intelligence, focusing on robotics and machine learning.

News

26/11/2018: I wrote an article (in German) for non-tech people, explaining the key ideas behind algorithms, machine learning and the limits of artificial intelligence. It's been officially published in JUSletter (requires membership, download the preprint here).

14/06/2018: Just in time for the World Cup: our paper on No free lunch in ball catching: A comparison of Cartesian and angular representations for control sheds some new light on 50 years of scientific debate about the gaze heuristic. Starring: (non-)optimal heuristics and (non-)optimal optimal control.

10/05/2018: Good things come to those who wait: the extended journal version of our Amazon Picking Challenge paper has been published in Autonomous Robots.

Projects

An excerpt of the projects that I completed in the last years.

Intuitive Machine Intelligence

I'm passionate about teaching non-tech people about algorithms, machine learning and artificial intelligence. Check out my latest introductory paper (German only) on this topic, as well as my blog on teaching machine learning and artificial intelligence without formal math.

Multi-Task Reinforcement Learning

Research paper on learning state representations in multi-task reinforcement learning.

The paper has been presented at the NIPS Workshop on Deep Learning for Action and Interaction. The spotlight talk is online on youtube.

PhD thesis: On Decomposability in Robot Reinforcement Learning

In my thesis, I studied two problems in robot learning, ball catching and learning manipulate articulated object, and I developed a conceptual framework for understanding solutions to these problems: the spectrum of decomposability.

CAPEL

Research paper on coupled learning of relational forward models and action parameters, presented at IROS 2016.

Amazon Picking Challenge 2015

In May 2015, our Team RBO won a prestigious international robotics challenge.

The lessons learned paper won the Best Systems Paper award at RSS 2016; the paper on the perception system got nominated for Best Paper Award at IROS 2016 (see here for extended journal version).

concarne

Lightweight ML framework for learning with side information (aka privileged information), based on Theano and Lasagne; also see our paper Patterns for Learning with Side Information.

Contact

Email me at mail - at - firstnamelastname - dot - de.

Disclaimer: All opinions and work presented here are my own and do not reflect the opinion of my employer.

Publications

Sebastian Höfer. Algorithmen, maschinelles Lernen und die Grenzen der KI. Jusletter 26. November 2018.

Sebastian Höfer and Jörg Raisch and Marc Toussaint and Oliver Brock. No free lunch in ball catching: A comparison of Cartesian and angular representations for control. PLoS ONE 13(6): e0197803, 2018.

Clemens Eppner^* and Sebastian Höfer^* and Rico Jonschkowski^* and Roberto Martín-Martín^* and Arne Sieverling^* and Vincent Wall^* and Oliver Brock. Four aspects of building robotic systems: lessons from the Amazon Picking Challenge 2015. Autonomous Robots, 2018.

Manuel Baum^*, Matthew Bernstein^*, Roberto Martín-Martín^*, Sebastian Höfer, Johannes Kulick, Marc Toussaint, Alex Kacelnik, Oliver Brock. Opening a Lockbox through Physical Exploration. Humanoids, Birmingham, UK, November 2017.

Sebastian Höfer. On decomposability in robot reinforcement learning. Dissertation. Technische Universität Berlin, Germany, June 2017.

Antonin Raffin^*, Sebastian Höfer^*, Rico Jonschkowski, Oliver Brock, and Freek Stulp. Unsupervised Learning of State Representations for Multiple Tasks. NIPS Workshop on Deep Learning for Action and Interaction, Barcelona, Spain, December 2016.

Sebastian Höfer and Oliver Brock. Coupled Learning of Action Parameters and Forward Models for Manipulation. IEEE/RSJ International Conference on Intelligent Robots and Systems, 2016.

Rico Jonschkowski and Clemens Eppner^* and Sebastian Höfer^* and Roberto Martín-Martín^* and Oliver Brock. Probabilistic Multi-Class Segmentation for the Amazon Picking Challenge. IEEE/RSJ International Conference on Intelligent Robots and Systems, 2016.

Clemens Eppner^* and Sebastian Höfer^* and Rico Jonschkowski^* and Roberto Martín-Martín^* and Arne Sieverling^* and Vincent Wall^* and Oliver Brock. Lessons from the Amazon Picking Challenge: Four Aspects of Building Robotic Systems. Proceedings of Robotics: Science and Systems, 2016.

Roberto Martín-Martín and Sebastian Höfer and Oliver Brock. An Integrated Approach to Visual Perception of Articulated Objects. Proceedings of the IEEE International Conference on Robotics and Automation, pp. 5091 - 5097, 2016.

Rico Jonschkowski^* and Sebastian Höfer^* and Oliver Brock. Patterns for Learning with Side Information. arXiv:1511.06429 [cs.LG] : 2016.

Marcus Buckmann and Robert Gaschler and Sebastian Höfer and Dennis Loeben and Peter A. Frensch and Oliver Brock. Learning to Explore the Structure of Kinematic Objects in a Virtual Environment. Frontiers in Psychology 6(374): 2015.

Malte Lorbach and Sebastian Höfer and Oliver Brock. Prior-Assisted Propagation of Spatial Information for Object Search. IEEE/RSJ International Conference on Intelligent Robots and Systems, pp. 2904-2909, 2014.

Sebastian Höfer and Tobias Lang and Oliver Brock. Extracting Kinematic Background Knowledge from Interactions Using Task-Sensitive Relational Learning. Proceedings of the IEEE International Conference on Robotics and Automation (ICRA), pp. 4342-4347, 2014.

Sebastian Höfer and Oliver Brock. Learning Compact Relational Models for the Exploration of Articulated Objects. Proceedings of the ICRA Mobile Manipulation Workshop on Interactive Perception (ICRA), 2013.

Luc Steels and Michael Spranger and Remi van Trijp and Sebastian Höfer and Manfred Hild. Emergent Action Language on Real Robots. Language Grounding in Robots. Springer (US), chap. 13, 255-276, 2012.

Katrien Beuls and Luc Steels and Sebastian Höfer. The Emergence of Internal Agreement Systems. Experiments in Cultural Language Evolution, pp. 233-256, 2012.

Sebastian Höfer and Michael Spranger and Manfred Hild. Posture Recognition Based on Slow Feature Analysis. Language Grounding in Robots. Springer Verlag, chap. 06, 111-130, 2012.

Katrien Beuls and Sebastian Höfer. Simulating the Emergence of Grammatical Agreement in Multi-agent Language Games. Twenty-Second International Joint Conference on Artificial Intelligence, pp. 61-66, 2011.

Sebastian Höfer. Anwendungen der Slow Feature Analysis in der humanoiden Robotik. Diploma Thesis, Humboldt University of Berlin, Germany, 2011.

Sebastian Höfer and Manfred Hild and Matthias Kubisch. Using Slow Feature Analysis to Extract Behavioural Manifolds Related to Humanoid Robot Postures. Tenth International Conference on Epigenetic Robotics, pp. 43-50, 2010.

Sebastian Höfer and Manfred Hild. Using Slow Feature Analysis to Improve the Reactivity of a Humanoid Robot's Sensorimotor Gait Pattern. International Conference on Neural Computation, pp. 212-219, 2010.

Michael Spranger and Sebastian Höfer and Manfred Hild. Biologically Inspired Posture Recognition and Posture Change Detection for Humanoid Robots. IEEE International Conference on Robotics and Biomimetics, pp. 562-567, 2009.

^* = shared first authorship.