Loading presentation...

Present Remotely

Send the link below via email or IM


Present to your audience

Start remote presentation

  • Invited audience members will follow you as you navigate and present
  • People invited to a presentation do not need a Prezi account
  • This link expires 10 minutes after you close the presentation
  • A maximum of 30 users can follow your presentation
  • Learn more about this feature in our knowledge base article

Do you really want to delete this prezi?

Neither you, nor the coeditors you shared it with will be able to recover it again.


Untitled Prezi

No description

Jenab Amin

on 17 September 2013

Comments (0)

Please log in to add your comment.

Report abuse

Transcript of Untitled Prezi

Amin Dastanpour GS33326

As electronic commerce becomes more popular, the role of automated negotiation systems is expected to increase.
Recently, with the development of artificial intelligence and the technique of agent, it becomes a hotspot to apply agent into electronic commerce. Because agent has some attributes such as automation and sociability.
In the negotiation, the information about opponent is often incompletion and the process of negotiation usually is uncertainty and unsteadiness, so we import the learning mechanism into the negotiation, and the agent can coordinate its own action through learning, and can finish negotiation efficiently. In the paper, we apply the Q-learning into the multi-agent automated negotiation, and build a multi-agent automated negotiation model which has the learning mechanism.
Negotiation protocol
Negotiation Protocol describes the message flow among the negotiation entities, and presents the action restrictions. And also it is the action rule which the agents must abide by in the negotiation
The flow of negotiation
At the beginning, the seller agent weighs the profit and chooses an optimal offer to the buyer agent. After received the offer from the seller agent, the buyer agent updates recent belief according to opponent offer, and then estimates the offer to judge whether accepts it or not. If the buyer agent doesn’t accept the offer, then the buyer agent will adjusts its own strategy and proposes a new offer, and sends the offer to the seller agent
Reinforcement learning
in automated negotiation
In the negotiation, agents will obtain corresponding rewards only when the negotiation succeeds. According to reinforcement learning, the offer should be consider as an action that the agent takes to transfer its own state under center state, e.g. the state of agent is s(t) when the agent is sending the tth offer, and the offer is offer(t)(i.e. the action of the agent), and then the state becomes s(t+1) after the agent sends the offer to its opponent.
Reinforcement learning
in automated negotiation
We can see that the reward is lag when the agent takes action in one state. When negotiation is not finish, the agent immediate rewards is r(s(t),Offer(t))=0. Only the negotiation is successful, the agent can obtain positive rewards, otherwise, the agent will obtain negative rewards. So the Q value based on Q-reinforcement learning can be computed as follows:
The BMN system has been implemented in C#. A number of experiments have been performed to test the feasibility of the presented approach and to demonstrate how the agents can negotiation with different negotiation strategies. Below are two examples of the results of the negotiation with Q-learning and the negotiation without learning.
we mainly research the automated negotiation in the Multi-agent system, and import the reinforcement learning into the negotiation. Through analyzing the negotiation protocol and the negotiation flow, we put forward an open, dynamic and supporting learning model. And we focus on how the agent using Q-learning to propose offers in the negotiation and depict it in E-commerce.
Full transcript