MagNet uses neural networks to predict multi-agent dynamics from observations.
problem Predicting the evolution of complex multi-agent systems.
method Formulated a coupled non-linear network with ODE-based state evolution, trained a neural network to discover dynamics from observations.
result Orders of magnitude improvement in prediction accuracy over traditional models.
Neural Architecture Search has shown potential to automate the design of neural networks. Deep Reinforcement Learning based agents can learn complex architectural patterns, as well as explore a vast and compositional search space. On the other hand, evolutionary algorithms offer higher sample efficiency, which is criti…
NeuroMAS treats multi-agent systems as neural networks for scalable, trainable coordination.
problem Designing multi-agent systems as hand-designed workflows is inefficient and inflexible.
method NeuroMAS treats multi-agent systems as a neural network architecture with reinforcement learning for scalable coordination.
result NeuroMAS improves significantly over multi-agent baselines and can be scaled progressively.
The Neural Testbed evaluates joint predictions of neural agents, revealing their limitations.
problem Evaluating the quality of joint predictions generated by neural agents.
method Developed an open-source benchmark (The Neural Testbed) to assess agents' marginal and joint predictions.
result Popular Bayesian deep learning agents perform poorly on joint predictions, even with accurate marginal predictions.
MPLP learns neural network weights by treating operations as message-passing agents.
problem Training neural networks using gradient-based methods.
method MPLP abstracts neural network operations as message-passing agents, updating internal states and passing messages.
result MPLP outperforms traditional gradient-based methods on simple feed-forward neural networks.
We propose a lifelong learning architecture, the Neural Computer Agent (NCA), where a Reinforcement Learning agent is paired with a predictive model of the environment learned by a Differentiable Neural Computer (DNC). The agent and DNC model are trained in conjunction iteratively. The agent improves its policy in simu…
Neural networks improve scalability for agent-based modeling demonstrations.
problem Scalability issues in training models of dynamic systems from demonstrations.
method Use of neural networks to reduce the search space for agent-level parameters.
result More scalable architecture for reproducing emergent behavior from demonstrations.
Cross-dimensional neural networks improve AI in Catan game.
problem Challenging to build AI agents for Catan game using RL.
method Introduced cross-dimensional neural networks to handle game complexities.
result RL agent outperforms best heuristic agent in Catan.
One-shot path planning for multiple agents using neural networks.
problem Efficiently generating optimal or near-optimal paths for multiple agents in robotics.
method Utilizes fully convolutional neural networks for one-shot multi-agent path planning.
result Demonstrates successful generation of optimal or near-optimal paths in over 85% of cases for multi-path planning.
I2C enables agents to learn efficient communication without redundancy.
problem Redundant broadcast communication in multi-agent cooperation.
method I2C learns a prior for agent-agent communication via causal inference and reinforcement learning.
result I2C reduces communication overhead and improves multi-agent cooperative performance.
Graph neural networks help AI agents learn more complex language.
problem Understanding how AI agents learn and use language.
method Developed graph referential games to compare different AI models.
result Graph neural networks enable AI to learn more complex, compositional language.
AgentNet is a graph neural network that learns to walk graphs intelligently, outperforming traditional methods.
problem Graph-level tasks, especially distinguishing and classifying graphs.
method AgentNet uses a computational model inspired by sublinear algorithms, where neural agents walk the graph and collectively decide the output.
result AgentNet can distinguish and separate graphs that are hard to distinguish, outperforming traditional graph neural networks.
Proposes a graph neural network for efficient multi-agent routing.
problem Routing multiple agents in complex, sparsely connected graphs with dynamic traffic.
method Value Iteration Network based on graph neural networks for online coordination.
result Significantly outperforms traditional methods in terms of cost and runtime.
Hybrid model simulates market dynamics using neural stochastic background traders.
problem Lack of realistic LOB simulations that combine historical data and dynamic interactions.
method Neural stochastic background trader trained on historical LOB data, embedded in multi-agent simulation.
result Hybrid model recreates stylised market facts and financial herding behaviors.
Improves neural relational inference for dynamic multi-agent trajectories.
problem Limited accuracy of NRI in short output sequences for relational inference in multi-agent trajectories.
method Proposes DYnamic multi-AgentRelational Inference (DYARI) model to handle changing interactions over time.
result DYARI model outperforms NRI in dynamic relational inference tasks.
Unified policy controls diverse agents through modular neural networks.
problem Learning control policies for various agent morphologies.
method Shared Modular Policies (SMP) with decentralized control and message passing.
result A single modular policy controls multiple agent morphologies.
Deep reinforcement learning methods attain super-human performance in a wide range of environments. Such methods are grossly inefficient, often taking orders of magnitudes more data than humans to achieve reasonable performance. We propose Neural Episodic Control: a deep reinforcement learning agent that is able to rap…
A novel framework for adaptive multi-agent communication in reinforcement learning.
problem Manual specification of communication structures in multi-agent reinforcement learning.
method Learning Structured Communication (LSC) framework using hierarchical graph neural networks.
result Adaptive hierarchical formations and efficient message propagation among agents.
ReF-ER algorithm improved performance in multi-agent reinforcement learning.
problem Improving performance in multi-agent reinforcement learning environments.
method Extended ReF-ER algorithm to include dependencies between agents and modeled environment dynamics.
result ReF-ER MARL outperforms state-of-the-art algorithms in collaborative environments.
VectorNet predicts car behavior using vectorized HD maps and agent dynamics.
problem Predicting behavior in multi-agent systems with self-driving cars.
method VectorNet uses hierarchical graph neural networks on vectorized representations of HD maps and agent trajectories.
result VectorNet achieves comparable or better performance than state-of-the-art methods while using fewer parameters and less computational power.
Recently, deep reinforcement learning (RL) methods have been applied successfully to multi-agent scenarios. Typically, these methods rely on a concatenation of agent states to represent the information content required for decentralized decision making. However, concatenation scales poorly to swarm systems with a large…
Model predicts multi-agent trajectories using a differentiable simulator.
problem Predicting future positions of multiple interacting agents.
method Conditional recurrent variational neural networks (CVRNNs) with a kinematic bicycle model.
result Achieves state-of-the-art results on INTERACTION dataset.
New approach for open ad hoc teamwork using graph-based policy learning.
problem Designing autonomous agents to collaborate with changing teams without prior coordination.
method Graph-based policy learning to adapt to dynamic team compositions.
result Successfully models the effects of other agents, leading to robust adaptation and superior performance.
Randomized neural networks improve deep RL agents' generalization.
problem Deep RL agents struggle to generalize to new, semantically similar environments.
method Introduce a randomized (convolutional) neural network to perturb input observations.
result Significantly outperforms various regularization and data augmentation methods.
Landmark2Vec maps unknown landmarks without GPS.
problem Estimate positions of unknown landmarks without GPS.
method Unsupervised neural network trained on landmark signals.
result Maps landmarks up to scale, rotation, and shift.
Reinforcement learning (RL) algorithms allow agents to learn skills and strategies to perform complex tasks without detailed instructions or expensive labelled training examples. That is, RL agents can learn, as we learn. Given the importance of learning in our intelligence, RL has been thought to be one of key compone…
Graph neural networks learn decentralized controllers from data.
problem Finding optimal decentralized controllers for autonomous agents is challenging.
method Adapting graph neural networks to handle delayed communications and ensure scalability and transferability.
result Graph neural networks can learn decentralized controllers from data, addressing the scalability and practical implementation issues of centralized controllers.
Deep neural net solves multi-agent optimal trading problem.
problem Optimal trade execution for multiple agents and assets.
method Residual U-net with self-attention for viscosity solution approximation.
result Neural network approach outperforms finite difference methods.
Fractal neural networks play SimCity and Conway's Game of Life on varying scales.
problem Generalizing agents' performance to larger gameboards than during training.
method Reinforcement learning in a custom environment, using fractal neural networks.
result Agents can generalize to larger gameboards, solving a minigame unsolvable with local strategies.
Deep active inference agents learn complex environments using Monte-Carlo methods.
problem Understanding and modeling biological intelligence in complex, continuous state-spaces.
method Neural architecture for deep active inference agents using multiple forms of Monte-Carlo sampling.
result Deep active inference agents can learn environmental dynamics and plan future actions.
Many potential applications of reinforcement learning in the real world involve interacting with other agents whose numbers vary over time. We propose new neural policy architectures for these multi-agent problems. In contrast to other methods of training an individual, discrete policy for each agent and then enforcing…
Meta-agent learns effective exploration from offline data.
problem Design a meta-agent to quickly maximize reward in unseen tasks.
method Bayesian RL approach with adaptive neural belief estimate.
result Meta-agent learns effective exploration behavior from diverse tasks.
With the breakthrough of computational power and deep neural networks, many areas that we haven't explore with various techniques that was researched rigorously in past is feasible. In this paper, we will walk through possible concepts to achieve robo-like trading or advising. In order to accomplish similar level of pe…
We propose a new perspective on adversarial attacks against deep reinforcement learning agents. Our main contribution is CopyCAT, a targeted attack able to consistently lure an agent into following an outsider's policy. It is pre-computed, therefore fast inferred, and could thus be usable in a real-time scenario. We sh…
ISMCTS-BR learns best responses in large games, approximating worst-case performance.
problem Learning robustness to worst-case outcomes in large games.
method ISMCTS-BR, a scalable search-based algorithm for deep reinforcement learning.
result ISMCTS-BR approximates worst-case performance in large games.
Graph neural network predicts vehicle interactions and trajectories for autonomous driving.
problem Predicting future motion of vehicles in traffic scenes.
method Graph neural network that jointly predicts interaction modes and 5-second future trajectories.
result Jointly predicting trajectories and interaction modes leads to lower trajectory error.
We introduce two approaches for combining neural evolution strategy (NES) and proximal policy optimization (PPO): parameter transfer and parameter space noise. Parameter transfer is a PPO agent with parameters transferred from a NES agent. Parameter space noise is to directly add noise to the PPO agent`s parameters. We…
Fast risk assessment for autonomous vehicles using learned agent futures.
problem Risk assessment for autonomous vehicles given probabilistic predictions of other agents' futures.
method Non-sampling based methods using deep neural networks for probabilistic predictions, with Gaussian and non-Gaussian mixture models for agent positions and controls.
result Effective risk assessment for low probability events using learned models of agent futures.
We tackle the blackbox issue of deep neural networks in the settings of reinforcement learning (RL) where neural agents learn towards maximizing reward gains in an uncontrollable way. Such learning approach is risky when the interacting environment includes an expanse of state space because it is then almost impossible…
We propose a unified mechanism for achieving coordination and communication in Multi-Agent Reinforcement Learning (MARL), through rewarding agents for having causal influence over other agents' actions. Causal influence is assessed using counterfactual reasoning. At each timestep, an agent simulates alternate actions t…
RH-UCRL combines pessimism and optimism for robust RL.
problem Ensuring reliable performance in real-world RL tasks with worst-case scenarios.
method RH-UCRL is a model-based RL algorithm that optimizes between an agent and an adversary, distinguishing between epistemic and aleatoric uncertainty.
result RH-UCRL achieves near-optimal sample complexity guarantees and outperforms other robust RL algorithms in adversarial environments.
We introduce a new RL problem where the agent is required to generalize to a previously-unseen environment characterized by a subtask graph which describes a set of subtasks and their dependencies. Unlike existing hierarchical multitask RL approaches that explicitly describe what the agent should do at a high level, ou…
Efficiently learns MFC systems with unknown dynamics.
problem Learning in multi-agent systems with non-stationary interactions and combinatorial state/action spaces.
method Model-based reinforcement learning algorithm, M3−UCRL, balancing exploration and exploitation. result First general regret bounds for model-based reinforcement learning of MFC systems.
GraSP-RL uses graph neural networks to improve job shop scheduling.
problem Capturing machine-unit-job sequence relationships and managing state space growth.
method Graph neural networks for feature extraction, reinforcement learning for decision-making, decentralized optimization.
result GraSP-RL outperforms existing methods in minimizing makespan for complex production environments.
The paper studies how grid cell patterns emerge in neural networks.
problem Understanding how grid cells in the brain form hexagonal firing patterns.
method Training recurrent neural networks with conformal normalization of velocity inputs.
result Conformal normalization is crucial for the emergence of hexagonal grid patterns in neural networks.
We explore neural painters, a generative model for brushstrokes learned from a real non-differentiable and non-deterministic painting program. We show that when training an agent to "paint" images using brushstrokes, using a differentiable neural painter leads to much faster convergence. We propose a method for encoura…
Assisted by neural networks, reinforcement learning agents have been able to solve increasingly complex tasks over the last years. The simulation environment in which the agents interact is an essential component in any reinforcement learning problem. The environment simulates the dynamics of the agents' world and henc…
FEPS models agents to learn and act in complex environments without deep neural networks.
problem Modeling complex adaptive systems and understanding self-organizing behavior.
method Introducing Free Energy Projective Simulation (FEPS) within the constraints of the free energy principle and active inference.
result FEPS agents resolve ambiguity and infer optimal policies in partially observable environments.