Existing multi-agent reinforcement learning methods are limited typically to a small number of agents. When the agent increases largely, becomes intractable due curse dimensionality and exponential growth interactions. In this paper, we present \emph{Mean Field Reinforcement Learning} where interactions within population agents approximated by those between single average effect from overall or neighboring agents; interplay two entities is mutually reinforced: individual agent's optimal...

A fundamental question in any peer-to-peer ridesharing system is how to, both effectively and efficiently, dispatch user's ride requests to the right driver real time. Traditional rule-based solutions usually work on a simplified problem setting, which requires sophisticated hand-crafted weight design for either centralized authority control or decentralized multi-agent scheduling systems. Although recent approaches have used reinforcement learning provide combinatorial optimization...

How to optimally dispatch orders vehicles and how trade off between immediate future returns are fundamental questions for a typical ride-hailing platform. We model as large-scale parallel ranking problem study the joint decision-making task of order dispatching fleet management in online platforms. This brings unique challenges following four aspects. First, facilitate huge number act learn efficiently robustly, we treat each region cell an agent build multi-agent reinforcement learning...

Coordination is one of the essential problems in multi-agent systems. Typically reinforcement learning (MARL) methods treat agents equally and goal to solve Markov game an arbitrary Nash equilibrium (NE) when multiple equilibra exist, thus lacking a solution for NE selection. In this paper, we unequally consider Stackelberg as potentially better convergence point than terms Pareto superiority, especially cooperative environments. Under games, formally define bi-level problem finding...

Effective fishway design requires knowledge of fish swimming behavior in streams and channels. Appropriate tests with near-natural flow conditions are required to assess the interaction between turbulent flows. In this study, volitional S. prenanti was tested quantified an open-channel flume three (low, moderate, high) regimes. The results showed that, when confronted alternative regimes, preferred select regions low velocities (0.25–0.50 m/s) kinetic energy (<0.05 m2/s2) for swimming,...

In this paper, we present C-ADAM, the first adaptive solver for compositional problems involving a non-linear functional nesting of expected values. We proof that C-ADAM converges to stationary point in $\mathcal{O}(δ^{-2.25})$ with $δ$ being precision parameter. Moreover, demonstrate importance our results by bridging, time, model-agnostic meta-learning (MAML) and optimisation showing fastest known rates deep network adaptation to-date. Finally, validate findings set experiments from...

This paper is concerned with multi-view reinforcement learning (MVRL), which allows for decision making when agents share common dynamics but adhere to different observation models. We define the MVRL framework by extending partially observable Markov processes (POMDPs) support more than one model and propose two solution methods through augmentation cross-view policy transfer. empirically evaluate our method demonstrate its effectiveness in a variety of environments. Specifically, we show...

A fundamental question in any peer-to-peer ridesharing system is how to, both effectively and efficiently, dispatch user's ride requests to the right driver real time. Traditional rule-based solutions usually work on a simplified problem setting, which requires sophisticated hand-crafted weight design for either centralized authority control or decentralized multi-agent scheduling systems. Although recent approaches have used reinforcement learning provide combinatorial optimization...

The primary objective of this investigation was to study the effect altitude on fish swimming ability. Different species were tested ensure that differences observed are not associated with a single species. Fish critical speed and burst determined using stepped-velocity tests in Brett-type respirometer. Based effects water temperature dissolved oxygen, it is clear ability decreases as increases. Further, because high physiology go beyond lower we recommend be at an similar target fishway...

Generative adversarial networks (GANs) have seen significant research interest over the past decade, yet core issues of training instability and mode collapse persist. This work proposes SwarmGAN, a novel GAN framework incorporating swarm intelligence to address these limitations. Specifically, exhibits properties well-suited enhance training: emergent complex behaviors arising from simple individual agents, decentralized adaptability instantaneous data hyperparameters, robustness through...

The operation of dams in the Yangtze River basin has historically induced severe total dissolved gas supersaturation (TDGS) due to flood discharges, adversely affecting swimming behavior resident fish. This disruption is compounded by habitat fragmentation resulting from numerous dam constructions, making fish passage facilities a critical mitigation strategy. study assessed performance bighead carp controlled vertical slot fishway under varied flow velocities 0.2, 0.25, and 0.3 m/s after...

<title>Abstract</title> During the flood season, high dam operations for discharge result in total dissolved gas (TDG) supersaturation. This condition causes bubble trauma (GBT) and can even lead to fish mortality, posing a significant threat downstream river ecosystems. Assessing ecological risks of TDG presents major challenge water power-intensive basins worldwide. Limited research has explored impact on behaviors such as aggression memory, which are crucial feeding, reproduction,...

Multi-agent reinforcement learning (MARL) has become effective in tackling discrete cooperative game scenarios. However, MARL yet to penetrate settings beyond those modelled by team and zero-sum games, confining it a small subset of multi-agent systems. In this paper, we introduce new generation learners that can handle nonzero-sum payoff structures continuous settings. particular, study the problem class games known as stochastic potential (SPGs) with state-action spaces. Unlike which all...

Trust region methods are widely applied in single-agent reinforcement learning problems due to their monotonic performance-improvement guarantee at every iteration. Nonetheless, when multi-agent settings, the of trust no longer holds because an agent's payoff is also affected by other agents' adaptive behaviors. To tackle this problem, we conduct a game-theoretical analysis policy space, and propose method (MATRL), which enables optimization for learning. Specifically, MATRL finds stable...

This paper presents SmartPartition, an efficient approach to partition large-scala natural graphs [5]. We design a new partitioning algorithm which can perceive graph layout locality improve the performance of computation. Experimental results demonstrate that SmartPartition achieve significant reduction in ingress time and execution for both real-world synthetic datasets.

Coordination is one of the essential problems in multi-agent systems. Typically reinforcement learning (MARL) methods treat agents equally and goal to solve Markov game an arbitrary Nash equilibrium (NE) when multiple equilibra exist, thus lacking a solution for NE selection. In this paper, we \emph{unequally} consider Stackelberg as potentially better convergence point than terms Pareto superiority, especially cooperative environments. Under games, formally define bi-level problem finding...