9. Optimizing and satisficing metrics 优化指标和满足指标(《MACHINE LEARNING YEARNING》翻译)

优化指标和满足指标

这是组合多个评估指标的另一种方法。

假设你同时关心算法的准确率和运行时间。你需要在下面三个分类器中进行选择:
这里写图片描述
这里如果将准确率和运行时间组合为单个评估指标会看起来不太自然,例如:

Accuracy0.5RunningTime

这里你可以替代为这样:首先,定义一个“可接受”的运行时间。例如我们说任何运行时间在100ms以内的算法都是可接受的。这里,运行时间就是一个“满足度量(satisficing metric)”——你的算法只需要在这个指标上表现地“足够好(good enough)”就行,在这个意义上它应该满足最多100ms。准确率是一个“优化度量(optimizing metric)”。

如果你正在权衡N个不同的标准,例如模型的二进制文件大小(这对于移动app很重要,因为用户不想要下载很大的程序),运行时间和准确率等。你可以考虑设置其中N-1个标准为“满意(satisficing)”指标,也就是说你只需要他们满足特定的值即可。然后将最后一个定义为“优化(optimizing)”指标。例如,为二进制文件大小和运行时间设定可接受的阈值,并尝试在这些约束条件下不断优化准确率。

作为最后一个例子,假定你正在构建一个硬件设备,该设备使用麦克风监听用户说出的某个特定的“唤醒语(wakeword)”,从而唤醒系统。例如:Amazon Echo监听“Alexa”;苹果Siri监听“Hey Siri”;Android监听“Okay Google”;或百度app监听“你好百度(Hello Baidu)”。你同时关心假正例的比率(the false positive rate, 即当没有人说唤醒语时系统唤醒的频率)和假反例的比率(the false negative rate, 即当有人说出唤醒语时系统没有唤醒的频率)。对这个系统表现性能的一个合理的目标是最小化假反例的比率(优化指标,optimizing metric),同时满足每24小时操作出现不超过一个假正例即可(满足指标,satisficing metric).

一旦你的团队对评估指标进行优化,他们将能够更快地取得进步。

  • 0
    点赞
  • 1
    收藏
    觉得还不错? 一键收藏
  • 1
    评论
Design efficient machine learning systems that give you more accurate results About This Book Gain an understanding of the machine learning design process Optimize machine learning systems for improved accuracy Understand common programming tools and techniques for machine learning Develop techniques and strategies for dealing with large amounts of data from a variety of sources Build models to solve unique tasks Who This Book Is For This book is for data scientists, scientists, or just the curious. To get the most out of this book, you will need to know some linear algebra and some Python, and have a basic knowledge of machine learning concepts. What You Will Learn Gain an understanding of the machine learning design process Optimize the error function of your machine learning system Understand the common programming patterns used in machine learning Discover optimizing techniques that will help you get the most from your data Find out how to design models uniquely suited to your task In Detail Machine learning is one of the fastest growing trends in modern computing. It has applications in a wide range of fields, including economics, the natural sciences, web development, and business modeling. In order to harness the power of these systems, it is essential that the practitioner develops a solid understanding of the underlying design principles. There are many reasons why machine learning models may not give accurate results. By looking at these systems from a design perspective, we gain a deeper understanding of the underlying algorithms and the optimisational methods that are available. This book will give you a solid foundation in the machine learning design process, and enable you to build customised machine learning models to solve unique problems. You may already know about, or have worked with, some of the off-the-shelf machine learning models for solving common problems such as spam detection or movie classification, but to begin solving more complex pr
Table of Contents 1 Why Machine Learning Strategy 2 How to use this book to help your team 3 Prerequisites and Notation 4 Scale drives machine learning progress 5 Your development and test sets 6 Your dev and test sets should come from the same distribution 7 How large do the dev/test sets need to be? 8 Establish a single-number evaluation metric for your team to optimize 9 Optimizing and satisficing metrics 10 Having a dev set and metric speeds up iterations 11 When to change dev/test sets and metrics 12 Takeaways: Setting up development and test sets 13 Build your first system quickly, then iterate 14 Error analysis: Look at dev set examples to evaluate ideas 15 Evaluating multiple ideas in parallel during error analysis 16 Cleaning up mislabeled dev and test set examples 17 If you have a large dev set, split it into two subsets, only one of which you look at 18 How big should the Eyeball and Blackbox dev sets be? 19 Takeaways: Basic error analysis 20 Bias and Variance: The two big sources of error 21 Examples of Bias and Variance 22 Comparing to the optimal error rate 23 Addressing Bias and Variance 24 Bias vs. Variance tradeoff 25 Techniques for reducing avoidable bias Page 3 Machine Learning Yearning-Draft Andrew Ng26 Error analysis on the training set 27 Techniques for reducing variance 28 Diagnosing bias and variance: Learning curves 29 Plotting training error 30 Interpreting learning curves: High bias 31 Interpreting learning curves: Other cases 32 Plotting learning curves 33 Why we compare to human-level performance 34 How to define human-level performance 35 Surpassing human-level performance 36 When you should train and test on different distributions 37 How to decide whether to use all your data 38 How to decide whether to include inconsistent data 39 Weighting data 40 Generalizing from the training set to the dev set 41 Addressing Bias and Variance 42 Addressing data mismatch 43 Artificial data synthesis 44 The Optimization Verification test 45 General form of Optimization Verification test 46 Reinforcement learning example 47 The rise of end-to-end learning 48 More end-to-end learning examples 49 Pros and cons of end-to-end learning 50 Learned sub-components 51 Directly learning rich outputs Page 4 Machine Learning Yearning-Draft Andrew Ng52 Error Analysis by Parts 53 Beyond supervised learning: What’s next? 54 Building a superhero team - Get your teammates to read this 55 Big picture 56 Credits
Machine Learning, Optimization, and Big Data: Second International Workshop, MOD 2016, Volterra, Italy, August 26-29, 2016, Revised Selected Papers (Lecture Notes in Computer Science) This book constitutes revised selected papers from the Second International Workshop on Machine Learning, Optimization, and Big Data, MOD 2016, held in Volterra, Italy, in August 2016. The 40 papers presented in this volume were carefully reviewed and selected from 97 submissions. These proceedings contain papers in the fields of Machine Learning, Computational Optimization and DataScience presenting a substantial array of ideas, technologies, algorithms, methods and applications. Table of Contents Chapter 1. Machine Learning: Multi-site Evidence-Based Best Practice Discovery Chapter 2. Data-Based Forest Management with Uncertainties and Multiple Objectives Chapter 3. Metabolic Circuit Design Automation by Multi-objective BioCAD Chapter 4. A Nash Equilibrium Approach to Metabolic Network Analysis Chapter 5. A Blocking Strategy for Ranking Features According to Probabilistic Relevance Chapter 6. A Scalable Biclustering Method for Heterogeneous Medical Data Chapter 7. Neural Learning of Heuristic Functions for General Game Playing Chapter 8. Comparing Hidden Markov Models and Long Short Term Memory Neural Networks for Learning Action Representations Chapter 9. Dynamic Multi-Objective Optimization with jMetal and Spark: A Case Study Chapter 10. Feature Selection via Co-regularized Sparse-Group Lasso Chapter 11. Economic Lot-Sizing Problem with Remanufacturing Option: Complexity and Algorithms Chapter 12. A Branch-and-Cut Algorithm for a Multi-item Inventory Distribution Problem Chapter 13. Adaptive Targeting in Online Advertisement: Models Based on Relative Influence of Factors Chapter 14. Design of Acoustic Metamaterials Through Nonlinear Programming Chapter 15. Driver Maneuvers Inference Through Machine Learning Chapter 16. A Systems Biology Approach for Unsupervised Clustering of High-Dimensional Data Chapter 17. Large-Scale Bandit Recommender System Chapter 18. Automatic Generation of Sitemaps Based on Navigation Systems Chapter 19. A Customer Relationship Management Case Study Based on Banking Data Chapter 20. Lagrangian Relaxation Bounds for a Production-Inventory-Routing Problem Chapter 21. Convergence Rate Evaluation of Derivative-Free Optimization Techniques Chapter 22. The Learnability of Business Rules Chapter 23. Dynamic Programming with Approximation Function for Nurse Scheduling Chapter 24. Breast Cancer's Microarray Data: Pattern Discovery Using Nonnegative Matrix Factorizations Chapter 25. Optimizing the Location of Helicopter Emergency Medical Service Operating Sites Chapter 26. An Enhanced Infra-Chromatic Bound for the Maximum Clique Problem Chapter 27. Cultural Ant Colony Optimization on GPUs for Travelling Salesman Problem Chapter 28. Combining Genetic Algorithm with the Multilevel Paradigm for the Maximum Constraint Satisfaction Problem Chapter 29. Implicit Location Sharing Detection in Social Media Turkish Text Messaging Chapter 30. Fuzzy Decision-Making of a Process for Quality Management Chapter 31. A Bayesian Network Profiler for Wildfire Arsonists Chapter 32. Learning Optimal Decision Lists as a Metaheuristic Search for Diagnosis of Parkinson's Disease Chapter 33. Hermes: A Distributed-Messaging Tool for NLP Chapter 34. Deep Learning for Classification of Dental Plaque Images Chapter 35. Multiscale Integration for Pattern Recognition in Neuroimaging Chapter 36. Game Theoretical Tools for Wing Design Chapter 37. Fastfood Elastic Net: Combining Variable Selection with Kernel Expansion Approximations Chapter 38. Big Data Analytics in a Public General Hospital Chapter 39. Inference of Gene Regulatory Network Based on Radial Basis Function Neural Network Chapter 40. Establishment of Optimal Control Strategy of Building-Integrated Photovoltaic Blind Slat Angle by Considering Interior Illuminance and Electricity Generation

“相关推荐”对你有帮助么?

  • 非常没帮助
  • 没帮助
  • 一般
  • 有帮助
  • 非常有帮助
提交
评论 1
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值