Explanipedia

D5RL: Diverse Datasets for Data-Driven Deep Reinforcement Learning Open

Rafael Rafailov, Kyle Hatch, Anikait Singh, Laura Smith, Aviral Kumar , et al. · 2024

Offline reinforcement learning algorithms hold the promise of enabling data-driven RL methods that do not require costly or dangerous real-world exploration and benefit from large pre-collected datasets. This in turn can facilitate real-wo…

Time-Symmetric Motion Maximizes Energy Efficiency in Fluid Open

Philip Ball · 2024

Researchers discovered a trick for dragging an object in a fluid with minimal effort, suggesting an optimal strategy for nanorobots.

‘Quantum plumbing’ at the nanoscale Open

Philip Ball · 2024

Nanofluidics could be used to purify water, generate energy and build nanoscale machines. But when water flows through a carbon nanotube, classical fluid mechanics breaks down, leading to puzzling experimental findings that researchers hav…

Feynman’s Reversed Sprinkler Puzzle Solved Open

Philip Ball · 2024

<i>Chemistry Nobel Prize:</i> Quantum Rules Provide Controllable Colors Open

Philip Ball · 2023

This year's Nobel Prize in Chemistry recognizes the development of quantum dots, particles whose size controls their color, making them useful for technologies such as displays.

Geometry and kinematics of the Middle to Late Miocene salt tectonics, central Egyptian Red Sea margin Open

Moamen Ali, Hemin Koyi, William Bosworth, Marco Ligi, Philip Ball , et al. · 2023

The Red Sea basin includes a thick Middle to Late Miocene evaporitic succession that underwent halokinesis and caused intensive reshaping of the seafloor and the development of salt-tectonic structures. However, the ge-ometry and kinematic…

Crystal Defects Interact to Form Intricate Structures Open

Philip Ball · 2023

that may affect the properties of a material.

Synthetic Experience Replay Open

Cong Lu, Philip Ball, Jack Parker-Holder · 2023

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,…

Four Walls Good, Two Walls Bad for Confined Cells Open

Philip Ball · 2023

Segregation of chromosomes in dividing cells

Efficient Online Reinforcement Learning with Offline Data Open

Philip Ball, Laura Smith, Ilya Kostrikov, Sergey Levine · 2023

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 expe…

Memories Become Chaotic before They Are Forgotten Open

Philip Ball · 2023

Red Sea rifting in central Egypt: constraints from the offshore Quseir province Open

Moamen Ali, Alessandro Decarlis, Marco Ligi, Philip Ball, William Bosworth , et al. · 2022

The formation of oceanic crust in the southern and central Red Sea is generally accepted to have started at c. 5 Ma. However, the nature of the crust in the northern Red Sea is still debated. This paper describes the rift architecture, dyn…

Simulations Using a Quantum Computer Show the Technology’s Current Limits Open

Philip Ball · 2022

Quantum circuits still can't outperform classical ones when simulating molecules.

Learning General World Models in a Handful of Reward-Free Deployments Open

Yingchen Xu, Jack Parker-Holder, Aldo Pacchiano, Philip Ball, Oleh Rybkin , et al. · 2022

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…

Frequency Comb Measures Quantum Interference Open

Philip Ball · 2022

A multiwavelength laser source known as a frequency comb provides a new technique for atom interferometry, potentially leading to new tests of fundamental physics.

Imperialism’s long shadow: the UK universities grappling with a colonial past Open

Philip Ball · 2022

Breakthrough Prize for the Physics of Quantum Information…and of Cells Open

Philip Ball · 2022

The work of this quartet supplied much of the theoretical basis for quantum information technologies, such

How Cells Move through Narrow Spaces Open

Philip Ball · 2022

Ultrafast Switch from a Bose-Einstein Condensate Open

Philip Ball · 2022

A subpicosecond optical switch demonstrated in a semiconductor material moves researchers a step closer to an all-optical computer.

Bayesian Generational Population-Based Training Open

Xingchen Wan, Cong Lu, Jack Parker-Holder, Philip Ball, Vu Nguyen , et al. · 2022

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 ar…

Stabilizing Off-Policy Deep Reinforcement Learning from Pixels Open

Edoardo Cetin, Philip Ball, Steve Roberts, Oya Çeliktutan · 2022

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 …

Same State, Different Task: Continual Reinforcement Learning without Interference Open

Samuel Kessler, Jack Parker-Holder, Philip Ball, Stefan Zohren, Stephen Roberts · 2022

Continual Learning (CL) considers the problem of training an agent sequentially on a set of tasks while seeking to retain performance on all previous tasks. A key challenge in CL is catastrophic forgetting, which arises when performance on…

Manipulating Objects Using Air Bubbles and Sound Waves Open

Philip Ball · 2022

Centimeter-scale objects in liquid can be manipulated using the mutual attraction of two arrays of air bubbles in the presence of sound waves.

Challenges and Opportunities in Offline Reinforcement Learning from Visual Observations Open

Cong Lu, Philip Ball, Tim G. J. Rudner, Jack Parker-Holder, Michael A. Osborne , et al. · 2022

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 obs…

Multiphoton Generator on a Chip Open

Philip Ball · 2022

Multiphoton Generator on a ChipA device for producing up to six photons in a single event could open new doors to quantum technologies.By Philip Ball P roducing entangled photons, whose properties are interdependent, is key to several quan…

Asymmetrical lithospheric necking of Red Sea rift  Open

Thamer Aldaajani, Hany Khalil, Philip Ball, Fabio A. Capitanio, Khalid A. Almalki · 2022

<p>The Red Sea rift exhibits two distinct rifting styles: in the north, the rifting is magma-poor, the crust is hyperextended and the lithospheric necking is asymmetric, in the south, rifting rapidly localized atop a symmetric lithos…

Geothermal energy as a means to decarbonize the energy mix of megacities Open

Carlos A. Vargas, Luca Caracciolo, Philip Ball · 2022

A new twist on graphene: an interview with Pablo Jarillo-Herrero and Allan MacDonald Open

Philip Ball · 2022

Graphene is the building block of graphite, made of carbon atoms bonded into sheets of hexagonal rings just a single atom thick. Although such isolated sheets had been predicted for many decades to exist, and had been grown on other surfac…

Ned Seeman (1945–2021) Open

Philip Ball · 2021

First quantum computer to pack 100 qubits enters crowded race Open

Philip Ball · 2021

Philip Ball YOU? Author Swipe