Filters








6,271 Hits in 5.3 sec

What is Going on Inside Recurrent Meta Reinforcement Learning Agents? [article]

Safa Alver, Doina Precup
2021 arXiv   pre-print
Recurrent meta reinforcement learning (meta-RL) agents are agents that employ a recurrent neural network (RNN) for the purpose of "learning a learning algorithm".  ...  Several illustrative experiments suggest that this hypothesis is true, and that recurrent meta-RL agents can be viewed as agents that learn to act optimally in partially observable environments consisting  ...  However, due to the black-box nature of neural networks, it is not clear what is going on inside the agent.  ... 
arXiv:2104.14644v1 fatcat:t6ch2mtrdvhuzkfr7orev6od4u

BADGER: Learning to (Learn [Learning Algorithms] through Multi-Agent Communication) [article]

Marek Rosa and Olga Afanasjeva and Simon Andersson and Joseph Davidson and Nicholas Guttenberg and Petr Hlubuček and Martin Poliak and Jaroslav Vítku and Jan Feyereisl
2019 arXiv   pre-print
Behavior, adaptation and learning to adapt emerges from the interactions of homogeneous experts inside a single agent.  ...  In this work, we propose a novel memory-based multi-agent meta-learning architecture and learning procedure that allows for learning of a shared communication policy that enables the emergence of rapid  ...  Multi-Agent Meta-Reinforcement Learning Existing work on MARL doesn't yet focus on meta-learning.  ... 
arXiv:1912.01513v1 fatcat:x2tczkviuvhnlirikx3w4tevy4

Autonomous Industrial Management via Reinforcement Learning: Self-Learning Agents for Decision-Making – A Review [article]

Leonardo A. Espinosa Leal, Magnus Westerlund, Anthony Chapman
2019 arXiv   pre-print
Once generalization is achieved, we discuss how these can be used to develop self-learning agents.  ...  We discuss using different types of extended realities, such as digital twins, to train reinforcement learning agents to learn specific tasks through generalization.  ...  One-shot visual imitation learning (OSVIL): Meta-imitation learning algorithm that teaches an agent to learn how to learn efficiently [59] .  ... 
arXiv:1910.08942v1 fatcat:wbpy3iijhbfwxeatiy4ztt5f2m

Deep Learning in (and of) Agent-Based Models: A Prospectus [article]

Sander van der Hoog
2017 arXiv   pre-print
A very timely issue for economic agent-based models (ABMs) is their empirical estimation.  ...  Economics has not yet benefited from these developments, and therefore we believe that now is the right time to apply Deep Learning and multi-layered neural networks to agent-based models in economics.  ...  Theme 4: Reinforcement learning in economic policy design The final theme is to apply a surrogate, or meta-modelling approach to policy decision-making.  ... 
arXiv:1706.06302v1 fatcat:gr5k3d3vnbc4jb43xwdnmzus5u

Deep Learning in Agent-Based Models: A Prospectus

Sander van der Hoog
2016 Social Science Research Network  
A very timely issue for economic agent-based models (ABMs) is their empirical estimation.  ...  Economics has not yet benefited from these developments, and therefore we believe that now is the right time to apply Deep Learning and multi-layered neural networks to agent-based models in economics.  ...  Financial support from the Horizon 2020 ISIGrowth Project (Innovation-fuelled, Sustainable, Inclusive Growth), under grant no. 649186, is gratefully acknowledged.  ... 
doi:10.2139/ssrn.2711216 fatcat:dvvofalherfyvi4pxpe54lbvne

SocialAI: Benchmarking Socio-Cognitive Abilities in Deep Reinforcement Learning Agents [article]

Grgur Kovač, Rémy Portelas, Katja Hofmann, Pierre-Yves Oudeyer
2021 arXiv   pre-print
Within the Deep Reinforcement Learning (DRL) field, this objective motivated multiple works on embodied language use.  ...  Building embodied autonomous agents capable of participating in social interactions with humans is one of the main challenges in AI.  ...  SocialEnv -In this meta-environment, which contains all previous ones, we consider a multi-task setup, in which the agent is facing a randomly drawn environment, i.e. it has to infer what is the current  ... 
arXiv:2107.00956v3 fatcat:6jyi3eivtfctbl2vl66se2jy3q

Building Intelligent Autonomous Navigation Agents [article]

Devendra Singh Chaplot
2021 arXiv   pre-print
The goal of this thesis is to make progress towards designing algorithms capable of 'physical intelligence', i.e. building intelligent autonomous navigation agents capable of learning to perform complex  ...  In the first part of the thesis, we discuss our work on short-term navigation using end-to-end reinforcement learning to tackle challenges such as obstacle avoidance, semantic perception, language grounding  ...  This motivates use of learning in a modular and hierarchical fashion inside of what one may call a 'classical navigation pipeline'.  ... 
arXiv:2106.13415v1 fatcat:5x5g64rd2rfvnmttcz5y7qvium

Observer effect from stateful resources in agent sensing

Adam Eck, Leen-Kiat Soh
2012 Autonomous Agents and Multi-Agent Systems  
In this model, the agent uses reinforcement learning to learn a controller for action selection, as well as how to predict expected knowledge refinement based on resource use during sensing.  ...  In this paper, we consider what happens when sensing actions require the use of stateful resources, which we define as resources whose state-dependent behavior changes over time based on usage.  ...  with what is believed to be full energy, or (2) reducing the learning rate to allow for additional exploration.  ... 
doi:10.1007/s10458-011-9189-y fatcat:dxlr5kjfxjgrhmekut5374y3bu

A Review of Cooperative Multi-Agent Deep Reinforcement Learning [article]

Afshin OroojlooyJadid, Davood Hajinezhad
2021 arXiv   pre-print
In this review article, we have focused on presenting recent approaches on Multi-Agent Reinforcement Learning (MARL) algorithms.  ...  Deep Reinforcement Learning has made significant progress in multi-agent systems in recent years.  ...  With the motivation of recent success on deep reinforcement learning (RL)-super-human level control on Atari games (Mnih et al. 2015) , mastering the game of Go , chess , robotic (Kober et al. 2013)  ... 
arXiv:1908.03963v4 fatcat:s2umqzxmqrhntkev3f6k554cv4

α-Rank: Multi-Agent Evaluation by Evolution [article]

Shayegan Omidshafiei, Christos Papadimitriou, Georgios Piliouras, Karl Tuyls, Mark Rowland, Jean-Baptiste Lespiau, Wojciech M. Czarnecki, Marc Lanctot, Julien Perolat, Remi Munos
2019 arXiv   pre-print
In contrast to the Nash equilibrium, which is a static concept based on fixed points, MCCs are a dynamical solution concept based on the Markov chain formalism, Conley's Fundamental Theorem of Dynamical  ...  We introduce α-Rank, a principled evolutionary dynamics methodology for the evaluation and ranking of agents in large-scale multi-agent interactions, grounded in a novel dynamical game-theoretic solution  ...  Specifically, PSRO can be viewed as a generalization of fictitious play, which computes approximate responses ("oracles") using deep reinforcement learning, along with arbitrary meta-strategy solvers;  ... 
arXiv:1903.01373v4 fatcat:p7qcpsvqsrctlhnocr5shaxfse

α-Rank: Multi-Agent Evaluation by Evolution

Shayegan Omidshafiei, Christos Papadimitriou, Georgios Piliouras, Karl Tuyls, Mark Rowland, Jean-Baptiste Lespiau, Wojciech M. Czarnecki, Marc Lanctot, Julien Perolat, Remi Munos
2019 Scientific Reports  
In contrast to the Nash equilibrium, which is a static solution concept based solely on fixed points, MCCs are a dynamical solution concept based on the Markov chain formalism, Conley's Fundamental Theorem  ...  of Dynamical Systems, and the core ingredients of dynamical systems: fixed points, recurrent sets, periodic orbits, and limit cycles.  ...  Specifically, PSRO can be viewed as a generalization of fictitious play, which computes approximate responses ("oracles") using deep reinforcement learning, along with arbitrary meta-strategy solvers;  ... 
doi:10.1038/s41598-019-45619-9 pmid:31289288 pmcid:PMC6617105 fatcat:25smi37gu5ey3lhzwfcgvda4wq

An actor-critic algorithm with deep double recurrent agents to solve the job shop scheduling problem [article]

Marta Monaci, Valerio Agasucci, Giorgio Grani
2021 arXiv   pre-print
We adopt an actor-critic scheme, where the action taken by the agent is influenced by policy considerations on the state-value function.  ...  The aim is to build up a greedy-like heuristic able to learn on some distribution of JSSP instances, different in the number of jobs and machines.  ...  Reinforcement Learning Reinforcement learning (RL) is a paradigm of machine learning, alongside supervised and unsupervised learning.  ... 
arXiv:2110.09076v1 fatcat:eymg3c37gbfjtparoyejwzoium

Learning Representations and Agents for Information Retrieval [article]

Rodrigo Nogueira
2019 arXiv   pre-print
A more efficient way is to train an artificial agent on how to use an external retrieval system to collect relevant information.  ...  A more limited, but still instrumental, version of this oracle is a question-answering system, in which an open-ended question is given to the machine, and an answer is produced based on the knowledge  ...  The agent is a neural network trained with supervised learning and fine-tuned with reinforcement learning.  ... 
arXiv:1908.06132v1 fatcat:xcujtsvsd5clljkteu4zrmr4hy

Social rules for agent systems [article]

René Mellema, Maarten Jensen, Frank Dignum
2021 arXiv   pre-print
Therefore there is hardly any research on how these social concepts relate and when each of them emerges or evolves from another concept.  ...  However, in the literature most papers concentrate on only one of these aspects at the time.  ...  This entails a learning mechanism inside the agents that keeps track how successful and how useful the social practice was.  ... 
arXiv:2004.12797v2 fatcat:ajo64rhd3fd6tbrz2nutk6bkey

Towards mental time travel: a hierarchical memory for reinforcement learning agents [article]

Andrew Kyle Lampinen, Stephanie C.Y. Chan, Andrea Banino, Felix Hill
2021 arXiv   pre-print
Agents with HCAM can extrapolate to task sequences much longer than they were trained on, and can even generalize zero-shot from a meta-learning setting to maintaining knowledge across episodes.  ...  Reinforcement learning agents often forget details of the past, especially after delays or distractor tasks.  ...  If we want Reinforcement Learning (RL) agents to meaningfully interact with complex environments over time, our agents will need to achieve this type of memory.  ... 
arXiv:2105.14039v3 fatcat:hwa33xqc5neexokyxobimmkgbe
« Previous Showing results 1 — 15 out of 6,271 results