ぱらぱらめくる『Statistical Relational Artificial Intelligence』

Statistical Relational Artificial Intelligence: Logic, Probability, and Computation (Synthesis Lectures on Artificial Intelligence and Machine Learning)

Statistical Relational Artificial Intelligence: Logic, Probability, and Computation (Synthesis Lectures on Artificial Intelligence and Machine Learning)

  • 作者: Luc De Raedt,Kristian Kersting,Sriraam Natarajan,David Poole
  • 出版社/メーカー: Morgan & Claypool
  • 発売日: 2016/03/24
  • メディア: ペーパーバック
  • この商品を含むブログを見る

  • StarAIって?:この本のMotivationから引用
the study and design of intelligent agents that act in worlds composed of individuals (objects, things), 

where there can be complex relations among the individuals, where the agents can be uncertain about what 

properties individuals have, what relations are true, what individuals exist, whether different terms denote 

the same individual, and the dynamics of the world.
  • 目次
    • 1 Motivation そもそも何が問題なのか、何がやりたいのか
    • Part I Representations どういう形式で表現するか
      • 2 Statistical and Relational AI Representations
      • 3 Relational Probabilistic Representations
      • 4 Representational Issues
    • Part II Inference 推定・推論
      • 5 Inference in Propositional Models
      • 6 Inference in Relational Probabilistic Models
    • Part III Learning 学習
      • 7 Learning Probabilistic and Logical Models
      • 8 Learning Probabilistic Relational Models
    • Part IV Beyond Probabilities 確率だけじゃない、その先は?
      • 9 Beyond Basid Probabilistic Inference and Learning
      • 10 Conclusions
  • 細目次
  • 1 Motivation
    • 1.1 Uncertainty in Complex Worlds 現実世界の不確実性の正体
    • 1.2 Challenges of Understanding StarAI 何がStarAIを理解することを難しくしているか
    • 1.3 The Benefits of Mastering StarAI どんなメリットがStarAIから得られるか
    • 1.4 Applications of StarAI 応用例
    • 1.5 Brief Historical Overview 成立の経緯
  • Part I Representation
  • 2 Statistical Relational AI Representations
    • 2.1 Probabilistic Graphical Models
      • ベイジアンネットワーク、マルコフネットワーク、ファクターグラフ
      • 有限個の固定個数の確率変数
    • 2.2 First-Order Logic and Logic Programming
      • 不定数の変数と論理式。論理関数が作る代数的世界
  • 3 Relational Probabilistic Representations
    • 前節の2つを組み合わせるということは、確率変数を持つ関数が作る代数にすること
    • 3.1 A General View: Parameterized Probabilistic Models
      • 確率変数の単純な扱いは、パラメタライズ分布関数として扱うこと
    • 3.2 Two Example Representations: Markov Logic And ProgLog
  • 4 Representational Issues
    • 4.1 Knowledge Representation Formalisms
      • 知識とは何か、それをどうやって表現するか(表現して計算機で扱うか)
    • 4.2 Objectives for Representation Language
      • 表現の仕方を決めるには色々な要請を満足させる(折り合いをつける)必要がある
        • Expresivity, Efficient inference, Understandability or explainability, Learnability, Compactness, Modularity, Ability to incorporate prior knowledge, Interoperability with heterogenous data, Latent variables
    • 4.3 Directed vs. Undirected models
      • 演算のやりかた、因果と結果、その他
    • 4.4 First-Order Logic vs. Logic Programs
      • 表せることに質的な違いがある
    • 以下の4.5-4.8は、変数の離散性とか確率(分布)をどう扱うかなどの課題
    • 4.5 Factors and Formulae
    • 4.6 Parameterizing Atoms
    • 4.7 Aggregators and Combining Rules
    • 4.8 Open Universe Models
  • Part II Inference
  • 5 Inference in Propositional Models 推定に2流派
    • 5.1 Probabilistic Inference
    • 5.2 Logical Inference
  • 6 Inference in Relational Probabilistic Models
    • 6.1 Grounded Inference for Relational Probabilistic Models
    • 6.2 Lifted Inference: Exploiting Symmetries
      • 術後論理を利用して確率を効率よく計算する
    • 6.3 (Lifted) Approximate Inference
  • Part III Learning
  • 7 Learning
    • 7.1 Learning Probabilistic Models
    • 7.2 Logical and Relational Learning
  • 8 Learning Probabilistic Relational Models
    • 8.1 Learning as Inference
    • 8.2 The Learning Problem
    • 8.3 Parameter Learning of Relational Models
    • 8.4 Structure Learning of Probabilistic Relational Models
    • 8.5 Bayesian Learning
  • ここまでは、論理と確率をどうやって組み合わせるかの話、それをするために使われる方法・表現法の話
  • Part IV Byeyond Probabilities
  • 9 Beyond Basic Probabilistic Inference and Learning
    • 9.1 Lifted Satisfiability
      • 充足可能性問題(SAT)があったが、それのLiftingに関わること
    • 9,2 Acting in Noisy Relational Worlds
    • 9.3 Relational Optimization
  • 10 Conclusions いくつかのポイント
    • Dynamics
    • Relational probabilistic modeling
    • Decision-theory and planning
    • Continuous distributions
    • Relational optimization
    • Approximate lifting and counting
    • Applications