欢迎光临澳大利亚新华书店网 [登录 | 免费注册]

    • 面向高风险应用的机器学习(影印版)(英文版)
      • 作者:(美)帕特里克·霍尔//詹姆士·柯蒂斯//帕鲁尔·潘迪|责编:张烨
      • 出版社:东南大学
      • ISBN:9787576612912
      • 出版日期:2024/03/01
      • 页数:438
    • 售价:55.2
  • 内容大纲

        过去十年人们见证了人工智能和机器学习(AI/ML)技术的广泛应用。然而,由于在广泛实施过程中缺乏监督,导致了一些本可以通过适当的风险管理来避免的事故和有害后果。在我们认识到AI/ML的真正好处之前,从业者必须了解如何降低其风险。
        本书描述了负责任的AI方法,这是一种以风险管理、网络安全、数据隐私、应用社会科学方面的最佳实践为基础,用于改进AI/ML技术、业务流程、文化能力的综合性框架。作者Patrick Hall、James Curtis、Parul Pandey为那些希望帮助组织、消费者和公众改善实际AI/ML系统成果的数据科学家创作了这本指南。
  • 作者介绍

  • 目录

    Foreword
    Preface
    Part Ⅰ. Theories and Practical Applications of AI Risk Management
      1.Contemporary Machine Learning Risk Management
        A Snapshot of the Legal and Regulatory Landscape
           The Proposed EU AI Act
           US Federal Laws and Regulations
           State and Municipal Laws
           Basic Product Liability
           Federal Trade Commission Enforcement
         Authoritative Best Practices
         AI Incidents
         Cultural Competencies for Machine Learning Risk Management
           Organizational Accountability
           Culture of Effective Challenge
           Diverse and Experienced Teams
           Drinking Our Own Champagne
           Moving Fast and Breaking Things
         Organizational Processes for Machine Learning Risk Management
           Forecasting Failure Modes
           Model Risk Management Processes
           Beyond Model Risk Management
         Case Study: The Rise and Fall of Zillow's iBuying ~
           Fallout
          Lessons Learned
        Resources
      2.Interpretable and Explainable Machine Learning
       Important Ideas for Interpretability and Explainability
       Explainable Models
          Additive Models
          Decision Trees
          An Ecosystem of Explainable Machine Learning Models
       Post Hoc Explanation
          Feature Attribution and Importance
          Surrogate Models
          Plots of Model Performance
          Cluster Profiling
       Stubborn Difficulties of Post Hoc Explanation in Practice
       Pairing Explainable Models and Post Hoc Explanation
       Case Study: Graded by Algorithm
       Resources
      3.Debugging Machine Learning Systems for Safety and Performance
       Training
          Reproducibility
          Data Quality
          Model Specification for Real-World Outcomes
       Model Debugging
          Software Testing
          Traditional Model Assessment
          Common Machine Learning Bugs

          Residual Analysis
          Sensitivity Analysis
          Benchmark Models
          Remediation: Fixing Bugs
       Deployment
         Domain Safety
         Model Monitoring
       Case Study: Death by Autonomous Vehicle
         Fallout
         An Unprepared Legal System
         Lessons Learned
       Resources
    ……
      4.Managing Bias in Machine Learning
      5.Security for Machine Learning
    Part Ⅱ.Putting AI Risk Management into Action
      6.Explainable Boosting Machines and Explaining XGBoost
      7.Explaining a PyTorch Image Classifier
      8.Selecting and Debugging XGBoost Models
      9.Debugging a PyTorch Image Classifier
      10.Testing and Remediating Bias with XGBoost
      11.Red-Teaming XGBoost
    Part Ⅲ.Conclusion
      12.How to Succeed in High-Risk Machine Learning