Zobrazeno 1 - 10
of 4 968
pro vyhledávání: '"Ball, A J"'
Autor:
Ball, Robert J.
Publikováno v:
Vergilius (1959-), 2024 Jan 01. 70, 99-130.
Externí odkaz:
https://www.jstor.org/stable/27339703
A key theme in the past decade has been that when large neural networks and large datasets combine they can produce remarkable results. In deep reinforcement learning (RL), this paradigm is commonly made possible through experience replay, whereby a
Externí odkaz:
http://arxiv.org/abs/2303.06614
Sample efficiency and exploration remain major challenges in online reinforcement learning (RL). A powerful approach that can be applied to address these issues is the inclusion of offline data, such as prior trajectories from a human expert or a sub
Externí odkaz:
http://arxiv.org/abs/2302.02948
Autor:
Ball, Olivier J.-P.1,2 (AUTHOR) olly.ball@wildlands.co.nz, Myers, Alan A.3 (AUTHOR) bavayia@gmail.com, Pohe, Stephen R.4 (AUTHOR) pohe.environmental@gmail.com, Shepherd, Lara D.5 (AUTHOR) lara.shepherd@tepapa.govt.nz
Publikováno v:
Diversity (14242818). Oct2024, Vol. 16 Issue 10, p632. 31p.
Autor:
Ball, Catie J., Haynes, Martha P., Jones, Michael G., Peng, Bo, Durbala, Adriana, Koopmann, Rebecca A., Ribaudo, Joseph, O'Donoghue, Aileen
The Baryonic Tully-Fisher Relation (BTFR) has applications in galaxy evolution as a testbed for the galaxy-halo connection and in observational cosmology as a redshift-independent secondary distance indicator. We use the 31,000+ galaxy ALFALFA sample
Externí odkaz:
http://arxiv.org/abs/2212.08728
Modelling the integrated H I spectra of galaxies has been a difficult task due to their diverse shapes, but more dynamical information is waiting to be explored in Hi line profiles. Based on simple assumptions, we construct a physically motivated mod
Externí odkaz:
http://arxiv.org/abs/2211.16455
Autor:
Xu, Yingchen, Parker-Holder, Jack, Pacchiano, Aldo, Ball, Philip J., Rybkin, Oleh, Roberts, Stephen J., Rocktäschel, Tim, Grefenstette, Edward
Building generally capable agents is a grand challenge for deep reinforcement learning (RL). To approach this challenge practically, we outline two key desiderata: 1) to facilitate generalization, exploration should be task agnostic; 2) to facilitate
Externí odkaz:
http://arxiv.org/abs/2210.12719
Autor:
Wan, Xingchen, Lu, Cong, Parker-Holder, Jack, Ball, Philip J., Nguyen, Vu, Ru, Binxin, Osborne, Michael A.
Reinforcement learning (RL) offers the potential for training generally capable agents that can interact autonomously in the real world. However, one key limitation is the brittleness of RL algorithms to core hyperparameters and network architecture
Externí odkaz:
http://arxiv.org/abs/2207.09405
Off-policy reinforcement learning (RL) from pixel observations is notoriously unstable. As a result, many successful algorithms must combine different domain-specific practices and auxiliary losses to learn meaningful behaviors in complex environment
Externí odkaz:
http://arxiv.org/abs/2207.00986
Autor:
Lu, Cong, Ball, Philip J., Rudner, Tim G. J., Parker-Holder, Jack, Osborne, Michael A., Teh, Yee Whye
Offline reinforcement learning has shown great promise in leveraging large pre-collected datasets for policy learning, allowing agents to forgo often-expensive online data collection. However, offline reinforcement learning from visual observations w
Externí odkaz:
http://arxiv.org/abs/2206.04779