Artificial Intelligence Programming with Python 笔记二

本章主要介绍了AI相关的开发工具，编程语言，框架等，属于比较简单的部分，略读即可，笔记如下：

关于硬件

Computers with GPUs For a more advanced user, if you want to train
more data or build more complicated, larger AI models, you will need
a more advanced computer, i.e., a computer with a graphics processing
unit (GPU). That typically includes a CPU with 8 cores, 32 GB RAM,
1 TB hard drive, and most importantly an NVIDIA GeForce RTX 1080 8GB GPU. GPUs, originally developed for video games, have
been increasingly used in machine learning and deep learning. GPUs can significantly speed up the calculations, thanks to their massive parallel processes. A prebuilt computer with a GPU is usually around $3,000, but you can get it much cheaper if you build it yourself.

带GPU的计算机如果你想训练，可以为更高级的用户。你需要更多的数据或构建更复杂、更大的人工智能模型

更先进的计算机，即具有图形处理功能的计算机单元（GPU）。这通常包括具有8个核的CPU、32GB RAM、32GB内存、32GB存储空间和32GB内存，1 TB硬盘，最重要的是NVIDIA GeForce RTX 1080 8GB GPU。

GPU最初是为视频游戏开发的，被越来越多地用于机器学习和深度学习。

GPU可以大大加快计算速度，这要归功于它们庞大的并行过程。一台预装GPU的电脑通常在3000美元左右，但如果你自己建造，你可以买到便宜得多的。

Computers with FPGAs Field Programmable Gate Array (FPGA) is also
increasingly being used in AI. Altera and Xilinx are the two most wellknown FPGA manufacturers. Acquiring Altera in 2015, Intel has since developed many FPGA-based AI applications. In 2018, for example, Intel achieved 3,700 frames per second processing with its Arria 10 GX 1150 FPGA. A computer with the Intel Arria 10 GX FPGA can easily cost more than $5,000

带有FPGA的计算机现场可编程门阵列（FPGA）也是Altera和Xilinx是两个最知名的FPGA制造商。

英特尔于2015年收购Altera，此后开发了许多基于FPGA的人工智能应用程序。例如，2018年，英特尔凭借其Arria 10 GX 1150 FPGA实现了每秒3700帧的处理。一台带有Intel Arria 10 GX FPGA的计算机的成本很容易超过5000美元。

简单的Python画图程序

import numpy as np
import matplotlib.pyplot as plt

x = np.linspace(-np.pi, np.pi, 50)
y1 = np.sin(x)
y2 = np.cos(x)
y3 = 2 * x + 1
y4 = x ** 3 + 9

plt.plot(x, y1, color = 'blue', marker = "s", label='Sin')
plt.plot(x, y2, color = 'red', marker = "o", label='Cos')
plt.plot(x, y3, color = 'green', marker = "o", label='2 * x + 1')
plt.plot(x, y4, color = 'yellow', marker = "o", label='x ** 3 + 9')

plt.legend()
plt.show()

AI 框架

1. Scikit-Learn is the most widely used Python framework for machine learning. It is divided into two main categories, supervised learning and unsupervised learning. For supervised learning, it has a set of algorithms for classification and regression.
2. Google’s TensorFlow is undoubtedly the most widely used Python framework for AI. It is called TensorFlow because its basic data structure is a tensor, a simple multidimensional array of numbers or functions. TensorFlow is widely used for image classification, object detection, image segmentation, pose detection, text sentiment, language translation, speech recognition, and so on.
3. Keras is an open source software library that provides a Python interface for other AI libraries and frameworks. Keras is best known as the standard interface for the TensorFlow. However, Keras, as a back end, also supports many others, such as Microsoft Cognitive Toolkit, R, Theano, and PlaidML.
4. Facebook’s PyTorch is an open source machine learning library based on the Torch library. It is developed by Facebook’s AI Research Lab. PyTorch is getting more popular in recent years and has caught up with TensorFlow.
5. OpenCV (Open Source Computer Vision Library) is probably the most popular cross-platform and free library. It was originally developed by Intel and aimed at real-time computer vision.

有价值的链接：
Dive into Deep Learning书籍：
Dive into Deep Learning