前言

最近注意到了NumSharp，想学习一下，最好的学习方式就是去实践，因此从github上找了一个用python实现的简单线性回归代码，然后基于NumSharp用C#进行了改写。

NumSharp简介

NumSharp（NumPy for C#）是一个在C#中实现的多维数组操作库，它的设计受到了Python中的NumPy库的启发。NumSharp提供了类似于NumPy的数组对象，以及对这些数组进行操作的丰富功能。它是一个开源项目，旨在为C#开发者提供在科学计算、数据分析和机器学习等领域进行高效数组处理的工具。

python代码

用到的python代码来源：llSourcell/linear_regression_live: This is the code for the “How to Do Linear Regression the Right Way” live session by Siraj Raval on Youtube (github.com)

下载到本地之后，如下图所示：

python代码如下所示：

#The optimal values of m and b can be actually calculated with way less effort than doing a linear regression. 
#this is just to demonstrate gradient descent
​
from numpy import *
​
# y = mx + b
# m is slope, b is y-intercept
def compute_error_for_line_given_points(b, m, points):
    totalError = 0
    for i in range(0, len(points)):
        x = points[i, 0]
        y = points[i, 1]
        totalError += (y - (m * x + b)) ** 2
    return totalError / float(len(points))
​
def step_gradient(b_current, m_current, points, learningRate):
    b_gradient = 0
    m_gradient = 0
    N = float(len(points))
    for i in range(0, len(points)):
        x = points[i, 0]
        y = points[i, 1]
        b_gradient += -(2/N) * (y - ((m_current * x) + b_current))
        m_gradient += -(2/N) * x * (y - ((m_current * x) + b_current))
    new_b = b_current - (learningRate * b_gradient)
    new_m = m_current - (learningRate * m_gradient)
    return [new_b, new_m]
​
def gradient_descent_runner(points, starting_b, starting_m, learning_rate, num_iterations):
    b = starting_b
    m = starting_m
    for i in range(num_iterations):
        b, m = step_gradient(b, m, array(points), learning_rate)
    return [b, m]
​
def run():
    points = genfromtxt("data.csv", delimiter=",")
    learning_rate = 0.0001
    initial_b = 0 # initial y-intercept guess
    initial_m = 0 # initial slope guess
    num_iterations = 1000
    print ("Starting gradient descent at b = {0}, m = {1}, error = {2}".format(initial_b, initial_m, compute_error_for_line_given_points(initial_b, initial_m, points)))
    print ("Running...")
    [b, m] = gradient_descent_runner(points, initial_b, initial_m, learning_rate, num_iterations)
    print ("After {0} iterations b = {1}, m = {2}, error = {3}".format(num_iterations, b, m, compute_error_for_line_given_points(b, m, points)))
​
if