Sam Blakeman
Senior Research Scientist
Email:
Address:
Luzern
Switzerland
Date of Birth:
06/06/1992
A Bit About Me
I am a Senior Research Scientist at Sony AI focusing on applying reinforcement learning to robotic control. The purpose of this site is to primarily share free tutorials on machine learning basics to help others start their journey in the field. These tutorials focus on key understandings that I wish I had reference to when first learning about the field. If you have any questions about them please do not hesitate to send me an email. On this site you will also find blog posts that reflect my own personal musings both about machine learning but also other topics close to my heart such as sport and general life topics.
My Background
Originally I studied natural sciences with a focus on biology and pharmacology. Through these studies I was exposed to Neuroscience at Yale University, which sparked my interest in the human mind and intelligence. Following on from my Bachelor degree I worked for a year as a software engineer, which was my first experience of programming and computer science. For my masters degree, I wanted to combine my interest in neuroscience and new-found love of computer science. I therefore did a masters in neuroscience with a focus on computational techniques at University College London. This lead to a PhD at Birkbeck, University of London focusing on efficient reinforcement learning in humans and machines. This marked a transition towards cognitive science and less of a focus on systems neuroscience. Following on from my PhD, I was motivated to apply my knowledge to a more practical problem as part of a wider team. This lead to me Sony AI where I am finding it extremely rewarding to investigate how we can use reinforcement learning for embodied agents in the real world.
Work Experience
2024 - present
2020 - 2024
2020 - 2020
2019 - 2020
2019 - 2019
2014 - 2015
2012 - 2013
Sony AI (CH) - Senior Research Scientist
Full time position as technical lead for the robotic control team. Responsible for taking the proof-of-concept from previous years and delivering the final solution for the end of the 5-year project. As a team of 4 we adapted, improved and scaled the model-free reinforcement learning control approach to achieve state-of-the-art results in the real-world. This culminated in a paper submission to Nature, which is currently under review.
Sony AI (CH) - Research Scientist
Full time position as a research scientist applying reinforcement learning to robotics. Part of a 5-year project combining high-speed sensing and robotic control with adversarial human interaction. Investigated the use of latent action spaces for model-free reinforcement learning on both simulated and real-world robotic arms. Demonstrated the benefit of such approaches over classical robotic control approaches such as Model Predictive Control and supervised learning.
Birkbeck, University of London (UK) - Research Associate
6-month EPSRC funding as part of the Human-Like Computing network. Using Deep Reinforcement Learning algorithms to explore human declarative memory and explainable AI.
The Alan Turing Institute (UK) - Research Assistant
Provided continued support for the London Air Quality project and Project Odysseus – Understanding London ‘Busyness’ and Exiting Lockdown. Primary role was to use computer vision techniques to extract data from London traffic cameras.
Imperial College London (UK) - Research Fellow, Data Science for Social Good
3-month internship using data science and machine learning to work on real world problems with a social impact. Worked with three other fellows to quantify traffic dynamics across greater London using video clips from traffic cameras and computer vision techniques.
Decent Group (UK) - Software Developer
Worked as part of a team of developers to produce bespoke database solutions (FileMaker) for small to medium sized businesses across a range of industries. Consulted with clients in person to outline project requirements and generate business flow charts.
Yale University (US) - Research Associate
1-year work placement investigating the role of nicotinic receptors in Major Depressive Disorder.
Publications
-
Blakeman, S., Mareschal D. (2022) Generating Explanations from Deep Reinforcement Learning Using Episodic Memory. arXiv
-
Blakeman, S., Mareschal, D. (2022) Selective Particle Attention: Rapidly and Flexibly Selecting Features for Deep Reinforcement Learning. Neural Networks.
-
Blakeman, S., Mareschal, D. (2020) A Complementary Learning Systems Approach to Temporal Difference Learning. Neural Networks.
-
Blakeman, S., Mareschal, D. (2017) Narrowing of the Cone-of-Direct Gaze Through Reinforcement Learning. Proceedings of the Annual Meeting of the Cognitive Science Society.
-
Mineur, Y.S. et al. [including Blakeman, S.] (2017) Hippocampal Nicotinic Ach Receptors Contribute to Modulation of Depression-Like Behaviour in C57BL/6J. British Journal of Pharmacology.
-
Mineur, Y.S, et al. [including Blakeman, S.] (2016) Multiple Nicotinic Acetylcholine Receptor Subtypes in the Mouse Amygdala Regulate Affective Behaviours and Response to Social Stress. Neuropsychopharmacology.
-
Mineur, et al. [including Blakeman, S.] (2015) Expression of the 5-HT1A Serotonin Receptor in the Hippocampus Is Required for Social Stress Resilience and the Antidepressant-Like Effects Induced by the Nicotinic Partial Agonist Cytisine. Neuropsychopharmacology.