Message Passing Multi-Agent GANs

نویسندگان

  • Arnab Ghosh
  • Viveka Kulharia
  • Vinay P. Namboodiri
چکیده

Communicating and sharing intelligence among agents is an important facet of achieving Artificial General Intelligence. As a first step towards this challenge, we introduce a novel framework for image generation: Message Passing Multi-Agent Generative Adversarial Networks (MPM GANs). While GANs have recently been shown to be very effective for image generation and other tasks, these networks have been limited to mostly single generator-discriminator networks. We show that we can obtain multi-agent GANs that communicate through message passing to achieve better image generation. The objectives of the individual agents in this framework are two fold: a co-operation objective and a competing objective. The co-operation objective ensures that the message sharing mechanism guides the other generator to generate better than itself while the competing objective encourages each generator to generate better than its counterpart. We analyze and visualize the messages that these GANs share among themselves in various scenarios. We quantitatively show that the message sharing formulation serves as a regularizer for the adversarial training. Qualitatively, we show that the different generators capture different traits of the underlying data distribution.

برای دانلود رایگان متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

منابع مشابه

Optimization of majority protocol for controlling transactions concurrency in distributed databases by multi-agent systems

In this paper, we propose a new concurrency control algorithm based on multi-agent systems which is an extension of majority protocol. Then, we suggest a clustering approach to get better results in reliability, decreasing message passing and algorithm’s runtime. Here, we consider n different transactions working on non-conflict data items. Considering execution efficiency of some different...

متن کامل

Efficient Agent Communication in Multi-agent Systems

In open multi-agent systems, agents are mobile and may leave or enter the system. This dynamicity results in two closely related agent communication problems, namely, efficient message passing and service agent discovery. This paper describes how these problems are addressed in the Actor Architecture (AA). Agents in AA obey the operational semantics of actors, and the architecture is designed t...

متن کامل

Coordination Approach to Find Best Defense Decision with Multiple Possibilities among Robocup Soccer Simulation Team

In 2D Soccer Simulation league, agents will decide based on information and data in their model. Effective decisions need to have world model information without any noise and missing data; however, there are few solutions to omit noise in world model data; so we should find efficient ways to reduce the effect of noise when making decisions. In this article we evaluate some simple solutions whe...

متن کامل

An Application of CoSMoS Design Methods to Pedestrian Simulation

In this paper, we discuss the implementation of a simple pedestrian simulation that uses a multi agent based design pattern developed by the CoSMoS research group. Given the nature of Multi Agent Systems (MAS), parallel processing techniques are inevitably used in their implementation. Most of these approaches rely on conventional parallel programming techniques, such as threads, Message Passin...

متن کامل

Parallel Fault Tolerant Multi-Agent Reinforcement Learning

Reinforcement learning is a powerful tool for training an agent in a sequential decision based environment and has been successful in many simulated [6] as well as practical [5] domains. In this paper we investigate methods of strengthening the rate of convergence of a single agent RL learner by sharing observations with other independent agents. In contrast to multi-agent reinforcement methods...

متن کامل

ذخیره در منابع من


  با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید

عنوان ژورنال:
  • CoRR

دوره abs/1612.01294  شماره 

صفحات  -

تاریخ انتشار 2016