《机器学习的理论困境与因果革命的七个启示》

目前几乎所有机器学习都基于统计方法，调整出更好的参数，拟合出更好的曲线。
但人类的优势在于知识表示、环境建模、自主假设，最终回答“如果……”一类的问题。

因果的三个层次与机器学习的理论困境

因果的三个层次

结构性因果模型（Structural Causal Models）与因果革命

  • 图形模型
  • 结构性方程
  • 反事实与干预逻辑

Structural Causal Models

因果革命中的七个支柱

-- 因果模型的独到之处

1. Encoding Causal Assumptions - Transparency & Testability

  • 透明：图形模型
  • 可测试：d-separation

2. The Control of Confounding

  1. back-door

back-door

  1. do-calculus

Do-Calculus

3. The formalization and Algorithmization of Counterfactuals

  • 图形表示
  • 结构性方程：右边因，左边果

4. Mediation Analysis & the Assessment of Direct and Indirect Effects

  • Mediation analysis
  • 图形表示

5. External Validity & Sample Selection Bias

  • Validity & Robustness
  • Association: Domain Adaptation, Transfer Learning, Life-long Learning, Explainable AI
  • Do-Calculus

6. Missing Data

  • Probabilistic Relationships

7. Causal Discovery

d-separation => Testability => { Infer, Prune } the models

Page Not Found

Try to search through the entire repo.