Python 语言构建机器学习系统 第2版

Preface
Chapter 1: Getting Started with Python Machine Learning
  Machine learning and Python - a dream team
  What the book will teach you (and what it will not)
  What to do when you are stuck
  Getting started
    Introduction to NumPy, SciPy, and matplotlib
    Installing Python
    Chewing data efficiently with NumPy and intelligentlywith SciPy
    Learning NumPy
      Indexing
      Handling nonexisting values
      Comparing the runtime
    Learning SciPy
 Our first (tiny) application of machine learning
    Reading in the data
    Preprocessing and cleaning the data
    Choosing the right model and learning algorithm
      Beforebuilding our first model...
      Starting with a simple straight line
      Towards some advanced stuff
      Stepping back to go forward - another look at our data
      Training and testing
      Answering our initial question
  Summary
Chapter 2: Classifying with Real-world Examples
  The Iris dataset
    Visualization is a good first step
    Building our first classification model
      Evaluation - holding out data and cross-validation
  Building more complex classifiers
  A more complex dataset and a more complex classifim
     Learning about the Seeds dataset
     Features and feature engineering
     Nearest neighbor classification
  Classifying with scikit-learn
     Looking at the decision boundaries
  Binary and multiclass classification
  Summary
Chapter 3: Clustering - Finding Related Posts
  Measuring the relatedness of posts
    How not to do it
    How to do it
  Preprocessing - similarity measured as a similar number of common words
    Converting raw text into a bag of words
      Counting words
      Normalizing word count vectors
      Removing less important words
      Stemming
      Stop words on steroids
    Our achievements and goals
  Clustering
    K-means
    Getting test data to evaluate our ideas on
    Clustering posts
  Solving our initial challenge
    Another look at noise
 Tweaking the parameters
 Summary
Chapter 4: Topic Modeling
  Latent Dirichlet allocation
    Building a topic model
  Comparing documents by topics
    Modeling the whole of Wikipedia
  Choosing the number of topics
  Summary
Chapter 5: Classification - Detecting Poor Answers
  Sketching our roadmap
  Learning to classify classy answers
    Tuning the instance
    Tuning the classifier
  Fetching the data
    Slimming the data down to chewable chunks
    Preselection and processing of attributes
    Defining what is a good answer
  Creating our first classifier
    Starting with kNN
    Engineering the features
    Training the classifier
    Measuring the classifier's performance
    Designing more features
  Deciding how to improve
    Bias-variance and their tradeoff
    Fixing high bias
    Fixing high variance
    High bias or low bias
  Using logistic regression
    A bit of math with a small example
    Applying logistic regression to our post classification problem
  Looking behind accuracy- precision and recall
  Slimming the classifier
  Ship it!
  Summary
Chapter 6: Classification II - Sentiment Analysis
  Sketching our roadmap
  Fetching the Twitter data
  Introducing the Naive Bayes classifier
    Getting to know the Bayes' theorem
    Being naive
    Using Naive Bayes to classify
    Accounting for unseen words and other oddities
    Accounting for arithmetic underflows
  Creating our first classifier and tuning it
    Solving an easy problem first
    Using all classes
    Tuning the classifier's parameters
  Cleaning tweets
  Taking the word types into account
    Determining the word types
    Successfully cheating using SentiWordNet
    Our first estimator
    Putting everything together
  Summary
Chapter 7: Regression
  Predicting house prices with regression
    Multidimensional regression
    Cross-validation for regression
  Penalized or regularized regression
    L1 and L2 penalties
    Using Lasso or ElasticNet in scikit-learn
    Visualizing the Lasso path
    P-greater-than-N scenarios
    An example based on text documents
    Setting hyperparameters in a principled way
  Summary
Chapter 8: Recommendations
  Rating predictions and recommendations
    Splitting into training and testing
    Normalizing the training data
    A neighborhood approach to recommendations
    A regression approach to recommendations
    Combining multiple methods
  Basket analysis
    Obtaining useful predictions
    Analyzing supermarket shopping baskets
    Association rule mining
    More advanced basket analysis
 Summary
Chapter 9: Classification - Music Genre Classification
  Sketching our roadmap
  Fetching the music data
    Converting into a WAV format
  Looking at music
    Decomposing music into sine wave components
  Using FFT to build our first classifier
    Increasing experimentation agility
    Training the classifier
    Using a confusion matrix to measure accuracy in
    multiclass problems
    An alternative way to measure classifier performance
    using receiver-operator characteristics
  Improving classification performance with Mel
  Frequency Cepstral Coefficients
  Summary
Chapter 10: Computer Vision
  Introducing image processing
    Loading and displaying images
    Thresholding
    Gaussian blurring
    Putting the center in focus
    Basic image classification
    Computing features from images
    Writing your own features
    Using features to find similar images
    Classifying a harder dataset
  Local feature representations
  Summary
Chapter 11: Dimensionality Reduction
  Sketching our roadmap
  Selecting features
    Detecting redundant features using filters
      Correlation
      Mutual information
    Asking the model about the features using wrappers
    Other feature selection methods
  Feature extraction
    About principal component analysis
      Sketching PCA
      Applying PCA
    Limitations of PCAand how LDA can help
  Multidimensional scaling
  Summary
Chapter 12: Bigger Data
  Learning about big data
    Using jug to break up your pipeline into tasks
    An introduction to tasks in jug
    Looking under the hood
    Using jug for data analysis
    Reusing partial results
  Using Amazon Web Services
    Creating your first virtual machines
      Installing Python packages on Amazon Linux
      Running jug on our cloud machine
    Automating the generation of clusters with StarCluster
  Summary
Appendix: Where to Learn More Machine Learning
  Online courses
  Books
    Question and answer sites
    Blogs
    Data sources
    Getting competitive
  All that was left out
  Summary
Index