Skip to content

Algorithmic Complex Computing-Agent Service

what 算法提供了什么服务

how valuable 有多大价值

1. Computing complexity and consequent transaction cost

2. Computing-Agent Service: to reduce the complexity and transaction cost

e.g. water sharing platform, no trade

  • no needs?
  • Pingzhong Tang (肾脏匹配)
  • apply an efficient algorithm -> many trade
  • Paper: Optimizing Trading Assignments in Water Right Markets
  • 经济学角度：理想自由市场 -> 最大化总福利

Why there's no trade

  • random communication & bargain -> bio distributed computing system -> it tries to solve the optimization problem above
  • trade only happens at the equilibrium; equilibrium is achieved by solving the computing problem by the "computers" (can be human)
  • Algorithm 1: random/natural communication -> brute-force searching -> \(O(N^2)\)
  • Algorithm 2: \(O(N)\)
  • Value of faster algorithms:
    • 需求有时效性
      • 如果不能更快求解，可能需求不存在了（天气：下雨周期 t ）
      • 可能市场状态在一段时间之后变化了，需要重新求解
      • 求解困难，需求以另一种方式满足了
    • Value of computing-agent service (计算代理服务)
      • 生活中有很多计算难问题：调度，分配，面部识别（记住/识别人是计算难的），制度设计（交易制度，政治制度）
      • 计算代理服务降低了「交易成本 transaction cost」（Coase）

3. Who needs the computing-agent service

  • 在哪部署计算代理服务收益最大，有价值：

    低速的自然模式计算带来了较高的交易成本

    计算建模 -> 发现计算负担重的任务 -> 在那里提供计算代理服务

    news vender problem; matching problem

  • transaction cost -> model it as an algorithm-design problem -> potential business opportunity for computing-agent service

    senario innovation 场景创新能力/方法论（而不是理论创新能力）

  • 2 types of algorithmic service industries

    • conventional consulting: computing agent (producer) & service demander (customer)

      买卖双方，单一市场；e.g. 云计算服务

    • equilibrium searching:

      两个市场：计算代理市场、需要均衡计算服务的交易市场

      e.g. 外卖市场：

      • 计算代理服务提供商：平台

      • 餐饮市场：商家、顾客、外卖配送员

        他们被平台「组织」起来工作；产生的附加问题：公司的边界

4. Value evaluation of computing-agent service

Welfare:

image-20220926200225996

How to evaluate the algorithm?

  • efficiency
  • accuracy

image-20220926200307330

We need to model the natural process to compare it with proposed algorithm.

  • Note that there're multiple ways to model a natural process.

What can affect the value of an algorithm?

From demand side,

  • environment, market settings
  • complexity of the problem
  • size of the problem

From supply side,

Other factors -> cost structure -> pricing in the market:

成本结构 -(决定)-> 市场中的竞争和定价

  • hardware (一次性)
  • system maintenance (持续投入)
  • energy cost (持续投入)

-> how to decide the price of an algorithm

Last update: September 26, 2022
Authors: Co1lin