Zobrazeno 1 - 10
of 63
pro vyhledávání: '"Hamrick, Jessica B."'
Autor:
Bounsi, Wilfried, Ibarz, Borja, Dudzik, Andrew, Hamrick, Jessica B., Markeeva, Larisa, Vitvitskyi, Alex, Pascanu, Razvan, Veličković, Petar
Transformers have revolutionized machine learning with their simple yet effective architecture. Pre-training Transformers on massive text datasets from the Internet has led to unmatched generalization for natural language understanding (NLU) tasks. H
Externí odkaz:
http://arxiv.org/abs/2406.09308
Autor:
Walker, Jacob, Vértes, Eszter, Li, Yazhe, Dulac-Arnold, Gabriel, Anand, Ankesh, Weber, Théophane, Hamrick, Jessica B.
State of the art reinforcement learning has enabled training agents on tasks of ever increasing complexity. However, the current paradigm tends to favor training agents from scratch on every new task or on collections of tasks with a view towards gen
Externí odkaz:
http://arxiv.org/abs/2302.04009
Autor:
Kosoy, Eliza, Chan, David M., Liu, Adrian, Collins, Jasmine, Kaufmann, Bryanna, Huang, Sandy Han, Hamrick, Jessica B., Canny, John, Ke, Nan Rosemary, Gopnik, Alison
Recent work in machine learning and cognitive science has suggested that understanding causal information is essential to the development of intelligence. The extensive literature in cognitive science using the ``blicket detector'' environment shows
Externí odkaz:
http://arxiv.org/abs/2206.08353
Autor:
Kosoy, Eliza, Liu, Adrian, Collins, Jasmine, Chan, David M, Hamrick, Jessica B, Ke, Nan Rosemary, Huang, Sandy H, Kaufmann, Bryanna, Canny, John, Gopnik, Alison
Despite recent progress in reinforcement learning (RL), RL algorithms for exploration still remain an active area of research. Existing methods often focus on state-based metrics, which do not consider the underlying causal structures of the environm
Externí odkaz:
http://arxiv.org/abs/2202.10430
Autor:
Anand, Ankesh, Walker, Jacob, Li, Yazhe, Vértes, Eszter, Schrittwieser, Julian, Ozair, Sherjil, Weber, Théophane, Hamrick, Jessica B.
One of the key promises of model-based reinforcement learning is the ability to generalize using an internal model of the world to make predictions in novel environments and tasks. However, the generalization ability of model-based agents is not well
Externí odkaz:
http://arxiv.org/abs/2111.01587
Autor:
Hamrick, Jessica B., Friesen, Abram L., Behbahani, Feryal, Guez, Arthur, Viola, Fabio, Witherspoon, Sims, Anthony, Thomas, Buesing, Lars, Veličković, Petar, Weber, Théophane
Model-based planning is often thought to be necessary for deep, careful reasoning and generalization in artificial agents. While recent successes of model-based reinforcement learning (MBRL) with deep function approximation have strengthened this hyp
Externí odkaz:
http://arxiv.org/abs/2011.04021
Autor:
Kosoy, Eliza, Collins, Jasmine, Chan, David M., Huang, Sandy, Pathak, Deepak, Agrawal, Pulkit, Canny, John, Gopnik, Alison, Hamrick, Jessica B.
Research in developmental psychology consistently shows that children explore the world thoroughly and efficiently and that this exploration allows them to learn. In turn, this early learning supports more robust generalization and intelligent behavi
Externí odkaz:
http://arxiv.org/abs/2005.02880
Autor:
Parascandolo, Giambattista, Buesing, Lars, Merel, Josh, Hasenclever, Leonard, Aslanides, John, Hamrick, Jessica B., Heess, Nicolas, Neitz, Alexander, Weber, Theophane
Standard planners for sequential decision making (including Monte Carlo planning, tree search, dynamic programming, etc.) are constrained by an implicit sequential planning assumption: The order in which a plan is constructed is the same in which it
Externí odkaz:
http://arxiv.org/abs/2004.11410
Autor:
Hamrick, Jessica B., Bapst, Victor, Sanchez-Gonzalez, Alvaro, Pfaff, Tobias, Weber, Theophane, Buesing, Lars, Battaglia, Peter W.
We introduce "Search with Amortized Value Estimates" (SAVE), an approach for combining model-free Q-learning with model-based Monte-Carlo Tree Search (MCTS). In SAVE, a learned prior over state-action values is used to guide MCTS, which estimates an
Externí odkaz:
http://arxiv.org/abs/1912.02807
Autor:
Bapst, Victor, Sanchez-Gonzalez, Alvaro, Shams, Omar, Stachenfeld, Kimberly, Battaglia, Peter W., Singh, Satinder, Hamrick, Jessica B.
We introduce agents that use object-oriented reasoning to consider alternate states of the world in order to more quickly find solutions to problems. Specifically, a hierarchical controller directs a low-level agent to behave as if objects in the sce
Externí odkaz:
http://arxiv.org/abs/1910.14361