NumPy

[NumPy] 是 Python 中科学计算的基础包。它是一个 Python 库，提供了多维数组对象、各种派生对象（如掩码数组和矩阵），以及一系列用于快速操作数组的例程，包括数学运算、逻辑运算、形状操作、排序、选择、输入输出、离散傅里叶变换、基本线性代数、基本统计操作、随机模拟等。

• 什么是 NumPy？
  • 为什么 NumPy 快？
  • 还有谁使用 NumPy？
• NumPy：初学者的绝对基础
  • 如何导入 NumPy
  • 为什么要使用 NumPy？
  • 什么是“数组”？
  • 数组基础
  • 数组属性
  • 如何创建基本数组
    • 指定数据类型
  • 添加、删除和排序元素
  • 如何知道数组的形状和大小？
  • 你可以 reshape 数组吗？
  • 如何将一维数组转换为二维数组（如何向数组添加新轴）
  • 索引和切片
  • 如何从现有数据创建数组
  • 基本数组操作
  • 广播（Broadcasting）
  • 更多有用的数组运算
  • 创建矩阵
  • 生成随机数
  • 如何获取唯一项及其计数
  • 转置和重塑矩阵
  • 如何反转数组
  • 重塑和展平多维数组
  • 如何访问文档字符串以获取更多信息
  • 如何使用数学公式
  • 如何保存和加载 NumPy 对象
• 数组创建
  • 介绍
  • 1) 将 Python 序列转换为 NumPy 数组
  • 2) 内置的 NumPy 数组创建函数
    • 1 - 一维数组创建函数
    • 2 - 二维数组创建函数
    • 3 - 通用多维数组创建函数
  • 3) 复制、连接或修改现有数组
  • 4) 从磁盘读取数组，无论是标准格式还是自定义格式
    • 标准二进制格式
    • 常见的 ASCII 格式
  • 5) 通过使用字符串或缓冲区从原始字节创建数组
  • 6) 使用特殊库函数（例如，SciPy、pandas 和 OpenCV）
• 对 ndarrays 的索引
  • 基本索引
    • 单元素索引
    • 切片和跨步
    • 维度索引工具
  • 高级索引
    • 整数数组索引
    • 布尔数组索引
    • 结合高级和基本索引
  • 字段访问
  • 扁平迭代器索引
  • 为索引数组赋值
  • 处理程序中可变数量的索引
  • 详细说明
• Importing data with genfromtxt()
  • Defining the input
  • Splitting the lines into columns
    • The delimiter argument
    • The autostrip argument
    • The comments argument
  • Skipping lines and choosing columns
    • The skip_header and skip_footer arguments
    • The usecols argument
  • Choosing the data type
  • Setting the names
    • The names argument
    • The defaultfmt argument
    • Validating names
  • Tweaking the conversion
    • The converters argument
    • Using missing and filling values
    • missing_values
    • filling_values
    • usemask
• 广播
  • 广播的一般规则
  • 可广播的数组
  • 一个实际示例：向量量化
• 拷贝与视图
  • 视图
  • 复制
  • 索引操作
  • 其他操作
  • 如何判断数组是视图还是副本
• Working with Arrays of Strings And Bytes
  • Fixed-width data types
  • Variable-width strings
    • Missing data support
      • NaN-like Missing Data Sentinels
      • String Missing Data Sentinels
      • Other Sentinels
    • Coercing Non-strings
    • Casting To and From Fixed-Width Strings
• Structured arrays
  • Introduction
  • Structured datatypes
    • Structured datatype creation
    • Manipulating and displaying structured datatypes
    • Automatic byte offsets and alignment
    • Field titles
    • Union types
  • Indexing and assignment to structured arrays
    • Assigning data to a structured array
      • Assignment from Python Native Types (Tuples)
      • Assignment from Scalars
      • Assignment from other Structured Arrays
      • Assignment involving subarrays
    • Indexing structured arrays
      • Accessing Individual Fields
      • Accessing Multiple Fields
      • Indexing with an Integer to get a Structured Scalar
    • Viewing structured arrays containing objects
    • Structure comparison and promotion
  • Record arrays
    • Recarray helper functions
      • append_fields()
      • apply_along_fields()
      • assign_fields_by_name()
      • drop_fields()
      • find_duplicates()
      • flatten_descr()
      • get_fieldstructure()
      • get_names()
      • get_names_flat()
      • join_by()
      • merge_arrays()
      • rec_append_fields()
      • rec_drop_fields()
      • rec_join()
      • recursive_fill_fields()
      • rename_fields()
      • repack_fields()
      • require_fields()
      • stack_arrays()
      • structured_to_unstructured()
      • unstructured_to_structured()
• Writing custom array containers
• 与 NumPy 的互操作性
  • 1. 在 NumPy 中使用任意对象
    • 数组接口协议
    • __array__() 方法
    • DLPack 协议
  • 2. 在不转换的情况下操作外部对象
    • __array_ufunc__ 协议
    • __array_function__ 协议
  • 3. 返回外部对象
  • 互操作性示例
    • 示例：Pandas 的 Series 对象
    • 示例：PyTorch 张量
    • 示例：CuPy 数组
    • 示例：Dask 数组
    • DLPack 示例
• Subclassing ndarray
  • Introduction
    • ndarrays and object creation
    • When to use subclassing
  • View casting
  • Creating new from template
  • Relationship of view casting and new-from-template
  • Implications for subclassing
    • A brief Python primer on __new__ and __init__
    • The role of __array_finalize__
  • Simple example - adding an extra attribute to ndarray
  • Slightly more realistic example - attribute added to existing array
  • __array_ufunc__ for ufuncs
  • __array_wrap__ for ufuncs and other functions
  • Extra gotchas - custom __del__ methods and ndarray.base
  • Subclassing and downstream compatibility
• NumPy 教程
  • Determining Moore’s Law with real data in NumPy
    • What you’ll do
    • Skills you’ll learn
    • What you’ll need
    • Building Moore’s law as an exponential function
    • Loading historical manufacturing data to your workspace
    • Calculating the historical growth curve for transistors
    • Sharing your results as zipped arrays and a csv
      • Zipping the arrays into a file
      • Creating your own comma separated value file
    • Wrapping up
    • References
  • Saving and sharing your NumPy arrays
    • What you’ll learn
    • What you’ll do
    • What you’ll need
    • Create your arrays
    • Save your arrays with NumPy’s savez
    • Remove the saved arrays and load them back with NumPy’s load
    • Reassign the NpzFile arrays to x and y
    • Success
    • Another option: saving to human-readable csv
    • Rearrange the data into a single 2D array
    • Save the data to csv file using savetxt
    • Our arrays as a csv file
    • Success, but remember your types
    • Wrapping up
  • Analyzing the impact of the lockdown on air quality in Delhi, India
    • What you’ll do
    • What you’ll learn
    • What you’ll need
    • The problem of air pollution
    • Building the dataset
    • Calculating the Air Quality Index
      • Moving averages
      • Sub-indices
      • Air quality indices
    • Paired Student’s t-test on the AQIs
      • Sampling
      • Defining the hypothesis
      • Calculating the test statistics
    • What do the t and p values mean?
    • In practice…
    • Further reading
  • Deep learning on MNIST
    • Prerequisites
    • Table of contents
    • 1. Load the MNIST dataset
    • 2. Preprocess the data
      • Convert the image data to the floating-point format
      • Convert the labels to floating point through categorical/one-hot encoding
    • 3. Build and train a small neural network from scratch
      • Neural network building blocks with NumPy
      • Model architecture and training summary
      • Compose the model and begin training and testing it
    • Next steps
  • Deep reinforcement learning with Pong from pixels
    • Prerequisites
    • Table of contents
      • A note on RL and deep RL
      • Deep RL glossary
    • Set up Pong
    • Preprocess frames (the observation)
    • Create the policy (the neural network) and the forward pass
    • Set up the update step (backpropagation)
    • Define the discounted rewards (expected return) function
    • Train the agent for a number of episodes
    • Next steps
    • Appendix
      • Notes on RL and deep RL
      • How to set up video playback in your Jupyter notebook
  • Masked Arrays
    • What you’ll do
    • What you’ll learn
    • What you’ll need
    • What are masked arrays?
    • When can they be useful?
    • Using masked arrays to see COVID-19 data
    • Exploring the data
    • Missing data
    • Fitting Data
    • In practice
    • Further reading
      • Reference
  • Plotting Fractals
    • What you’ll do
    • What you’ll learn
    • What you’ll need
    • Warmup
    • Julia set
    • Mandelbrot set
    • Generalizing the Julia set
      • Newton Fractals
    • Creating your own fractals
    • In conclusion
    • On your own
    • Further reading
  • Determining Static Equilibrium in NumPy
    • What you’ll do:
    • What you’ll learn:
    • What you’ll need:
    • Solving equilibrium with Newton’s second law
    • Solving Equilibrium as a sum of moments
    • Finding values with physical properties
      • Another Example
    • Wrapping up
      • Additional Applications
      • References
  • Linear algebra on n-dimensional arrays
    • Prerequisites
    • Learner profile
    • Learning Objectives
    • Content
      • Shape, axis and array properties
      • Operations on an axis
    • Approximation
      • Applying to all colors
      • Products with n-dimensional arrays
      • Final words
    • Further reading
  • X-ray image processing
    • Prerequisites
    • Table of contents
    • Examine an X-ray with imageio
    • Combine images into a multidimensional array to demonstrate progression
    • Edge detection using the Laplacian-Gaussian, Gaussian gradient, Sobel, and Canny filters
      • The Laplace filter with Gaussian second derivatives
      • The Gaussian gradient magnitude method
      • The Sobel-Feldman operator (the Sobel filter)
      • The Canny filter
    • Apply masks to X-rays with np.where()
    • Compare the results
    • Next steps