doi.bio/victor_bapst

Victor Bapst

Victor Bapst is a researcher in the field of AI and machine learning. He has worked with numerous co-authors on a variety of publications, primarily between 2013 and 2021.

Education

Bapst is associated with the Goethe-Universität Frankfurt am Main, Frankfurt am Main, Germany, where he works at the Institut für Mathematik.

Career

Bapst has worked with researchers from DeepMind, Google, the University of Oxford, and the University of Rome, among others.

Publications

Bapst's publications include:

• The Quantum Adiabatic Algorithm applied to random optimization problems: the quantum spin glass perspective (2012): Co-authored with Guilhem Semerjian and Francesco Zamponi.
• Planting colourings silently (2014): Co-authored with Amin Coja-Oghlan and Charilaos Efthymiou.
• The Condensation Phase Transition in Random Graph Coloring (2014): Co-authored with Amin Coja-Oghlan, Samuel Hetterich, Felicia Raßmann, and Dan Vilenchik.
• A positive temperature phase transition in random hypergraph 2-coloring (2014): Co-authored with Amin Coja-Oghlan.
• Harnessing the Bethe Free Energy (2015): Co-authored with Amin Coja-Oghlan.
• The condensation phase transition in the regular k-SAT model (2015): Co-authored with Amin Coja-Oghlan.
• Distral: Robust Multitask Reinforcement Learning (2017): Co-authored with Yee Whye Teh, Wojciech M. Czarnecki, John Quan, James Kirkpatrick, Raia Hadsell, Nicolas Heess, and Razvan Pascanu.
• Sample Efficient Actor-Critic with Experience Replay (2017): Co-authored with Ziyu Wang, Nicolas Heess, Volodymyr Mnih, Rémi Munos, Koray Kavukcuoglu, and Nando de Freitas.
• Structured agents for physical construction (2019): Co-authored with Alvaro Sanchez-Gonzalez, Carl Doersch, Kimberly L. Stachenfeld, Pushmeet Kohli, Peter W. Battaglia, and Jessica B. Hamrick.
• Combining Q-Learning and Search with Amortized Value Estimates (2019): Co-authored with Jessica B. Hamrick, Alvaro Sanchez-Gonzalez, Tobias Pfaff, Theophane Weber, Lars Buesing, and Peter W. Battaglia.

- A Deep Learning Approach for Characterizing Major Galaxy Mergers (2021): Co-authored with Skanda Koppula, Marc Huertas-Company, Sam Blackwell, Agnieszka Grabska-Barwinska, Sander Dieleman, and others.

Victor Bapst

Victor Bapst is a machine learning researcher with a background in mathematics and physics. He is currently affiliated with the Goethe-Universität Frankfurt am Main in Germany. Bapst has previously worked with DeepMind and the University of Southern California.

Education and Career

Bapst's educational background and career history are currently unclear. However, his research interests include combinatorial optimization, phase transitions in random graph coloring, and the application of quantum algorithms to optimization problems.

Publications

Bapst has authored or co-authored numerous research papers in the fields of machine learning, artificial intelligence, and computational mathematics. Notable publications include:

• The Condensation Phase Transition in the Regular k-SAT Model (2016): This paper, co-authored with Amin Coja-Oghlan, explores phase transitions in the context of the k-SAT problem, a fundamental challenge in computer science and discrete mathematics.
• Harnessing the Bethe Free Energy (2015,2016): Bapst and Coja-Oghlan propose a method for calculating the typical value of the partition function in Gibbs measures induced by random factor graphs, with applications in computer science, combinatorics, and physics.
• The Quantum Adiabatic Algorithm applied to random optimization problems: the quantum spin glass perspective (2012):** In this work, Bapst and colleagues investigate the use of quantum adiabatic algorithms for solving random optimization problems, providing insights into the efficiency of these algorithms compared to classical approaches.
• Distral: Robust Multitask Reinforcement Learning (2017):** Bapst, Teh, Czarnecki, and others propose a novel approach to multitask reinforcement learning, demonstrating its effectiveness in various tasks.
• Structured agents for physical construction (2019): Bapst and a team of researchers develop structured agents capable of learning to construct physical structures, drawing inspiration from relational inductive biases observed in human construction tasks.
• Hyperbolic Attention Networks (2018,2019):** This work introduces a new attention mechanism for neural networks, leveraging hyperbolic geometry to improve performance on tasks involving structured data, such as graphs and trees.
• Deep reinforcement learning with relational inductive biases (2018,2019):** Bapst and a large team of researchers combine deep reinforcement learning with relational inductive biases, demonstrating improved performance on tasks requiring relational reasoning, such as the game of StarCraft.

Co-authors

Bapst has collaborated with many prominent researchers in the field, including Peter W. Battaglia, Jessica B. Hamrick, Alvaro Sanchez-Gonzalez, Amin Coja-Oghlan, and others.

Victor Bapst

Victor Bapst is a machine learning researcher. He is currently affiliated with the Goethe-Universität Frankfurt am Main, Frankfurt am Main, Institut für Mathematik.

Education and Career

Bapst has studied and worked at several institutions, including:

• Goethe-Universität Frankfurt am Main
• Sapienza University of Rome
• Sorbonne Université
• University of Southern California
• Spanish National Research Council
• University of Naples Federico II
• Nordic Institute for Theoretical Physics
• Ben-Gurion University of the Negev
• Ecole Normale Supérieure de Paris
• Royal Holloway, University of London

Research Interests and Publications

Bapst's research interests include machine learning, reinforcement learning, graph theory, quantum computing, and statistical mechanics.

He has published extensively in these fields, with notable publications including:

• "The Quantum Adiabatic Algorithm applied to random optimization problems: the quantum spin glass perspective" (2012)
• "Planting Colourings Silently" (2014, 2017)
• "Distral: Robust Multitask Reinforcement Learning" (2017)
• "Structured agents for physical construction" (2019)
• "Combining Q-Learning and Search with Amortized Value Estimates" (2019, 2020)
• "Hyperbolic Attention Networks" (2018, 2019)
• "Deep reinforcement learning with relational inductive biases" (2018, 2019)
• "Relational inductive biases, deep learning, and graph networks" (2018)

Co-authors

Bapst has collaborated with many researchers in the field, including Peter W. Battaglia, Jessica B. Hamrick, Alvaro Sanchez-Gonzalez, Amin Coja-Oghlan, and others.