Stanford University Machine Learning assignment problem Set #1 regression for denoising Quasar spectra next article

(i) Processing of documents

#-*-Coding:utf-8-*-import numpy as NP import math import matplotlib.pyplot as PLT import csv def read (): Fr=open ( ' Quasar_test.csv ', ' R ') Arrayline=fr.readlines () Y=arrayline[1].strip (). Split (', ') M=len (arrayline) N=len (y ) y2 = Np.zeros ((m-1, n)) x = Np.zeros ((2, N)) x[0] = Arrayline[0].strip (). Split (', ') x[1] = 1 Y=map (Lambda y:float (y), y) y2[0] = weighted_linear_regression (x, Y, 5) q=[] for I in range (len (y2[0)): Q . Append (Y2[0][i]) f1 = open (' Result1.csv ', ' W ') f1.write (str (q)) f1.write (' \ n ') I=1 for line in Arrayl Ine[2:]: y = Line.strip (). Split (', ') y = map (lambda y:float (y), y) Y2[i]=weighted_linear_regressi On (x, Y, 5) q = [] for j in Range (Len (y2[0))): Q.append (Y2[i][j]) print Q i=i+
    1 f1.write (str (q)) f1.write (' \ n ') return y2 def weighted_linear_regression (x,y,t): #加权线性回归 y2=[] For I in range (Len (x[0])): W=np.zeros ((Len (x[0), Len (x[0))) for J in Range (Len (x[0)): W[j][j]=math.exp ((x[0][i)- X[0][J]) * (X[0][i]-x[0][j])/( -2 * t * t)) Xwx=np.dot (Np.dot (x,w), Np.transpose (x)) xwx_inverse=np.linal G.INV (XWX) Xwx_inverse_x=np.dot (xwx_inverse,x) xwx_inverse_x_w = Np.dot (xwx_inverse_x,w) xwx_inver Se_x_w_y=np.dot (Xwx_inverse_x_w,np.transpose (y)) theta=xwx_inverse_x_w_y y2.append (x[0][i] * theta[0] + th
 ETA[1]) return y2 Y2=read ()

(ii)

#-*-Coding:utf-8-*-import numpy as NP import HEAPQ import csv def read (): Fr=open (' result.csv ') Arrayline = f
    R.readlines () x = Arrayline[0].strip (). Split (', ') X=map (lambda x:float (x), X) M=len (arrayline) N=len (x) Y=np.zeros ((m-1,n)) i=0 for line in arrayline[1:: Num=line.strip (). Split (', ') num = map (lambd A num:float (num), num) y[i]=num i=i+1 return X,y x,y=read () def predict (x,y,m): dis=[] for I 
        In range (len (y)): sum = 0 to J in range (150,450): sum=sum+ (Y[m][j]-y[i][j]) * (Y[M][J]-Y[I][J))
    Dis.append (SUM) smallest_four=heapq.nsmallest (4, dis) smallest_three=smallest_four[1:] location=[]
    For I in range (len (smallest_three)): Location.append (Dis.index (smallest_three[i)) H=max (Smallest_three)
            F_left=[] for I in range (m): Top=0 bottom=0 for J in Range (Len (smallest_three)):
         N=LOCATION[J]   top = top + (1-smallest_three[j]/h) * y[n][i] bottom = bottom + (1-smallest_three[j]/h) sum=to  P/bottom f_left.append (sum) return f_left def error (Y): f_left_predict=[] sum_error=[] to M in Range (len (y)): error=0 f_left=predict (x, Y, M) for N in range (len (f_left)): Error=err  or+ (F_left[n]-y[m][n]) * (F_left[n]-y[m][n]) f_left_predict.append (f_left) sum_error.append (Error) Print SUM (sum_error)/len (sum_error) error (Y)

(iii)

#-*-Coding:utf-8-*-import numpy as NP import HEAPQ import Matplotlib.pyplot as Plt def read1 (): #读取数据 Fr=op
    En (' result.csv ') Arrayline = Fr.readlines () x = Arrayline[0].strip (). Split (', ') X=map (lambda x:float (x), X) M=len (Arrayline) N=len (x) Y=np.zeros ((m-1,n)) i=0 for line in arrayline[1:: Num=line.strip ().       Split (', ') num = map (lambda num:float (num), num) Y[i]=num i=i+1 return x,y def read2 (): #读取预测数据 fr=open (' result1.csv ') Arrayline = Fr.readlines () x= Arrayline[0].strip (). Split (', ') X=map (lambd
        A x:float (x), X) M=len (arrayline) N=len (x) Y=np.zeros ((m-1,n)) i=0 for line in Arrayline[1:]:

Num=line.strip (). Split (', ') num = map (lambda num:float (num), num) y[i]=num i=i+1 return x,y
            def predict (x,y,y1,m): #y1为预测数据 dis=[] for I in range (len (y)): sum = 0 for J in Range (150,450): sum=sum+ (y1[M][J]-Y[I][J]) * (Y1[m][j]-y[i][j]) dis.append (sum) smallest_three=heapq.nsmallest (3, dis) location=[] F or I in range (len (smallest_three)): Location.append (Dis.index (smallest_three[i)) H=max (Smallest_three) f_ Left=[] for I in range (m): Top=0 bottom=0 for J in Range (Len (smallest_three)): n= LOCATION[J] top = top + (1-smallest_three[j]/h) * y[n][i] bottom = bottom + (1-smallest_three[
    j]/h) sum=top/bottom f_left.append (sum) return F_left def error (X,Y,Y1): f_left_predict=[] Sum_error=[] for M in range (len (y1)): F_left=predict (x,y,y1,m) error = 0 to n in range (Len (f) _left)): error=error+ (F_left[n]-y1[m][n]) * (f_left[n]-y1[m][n) f_left_predict.append (f_left) s Um_error.append (Error) return F_left_predict[0],f_left_predict[5],sum (Sum_error)/len (sum_error) def figure (X,        EXAMPLE_1,EXAMPLE_6,Y1):#画图 plt.figure (1) plt.xlabel (' Wavelength ') plt.ylabel (' Flux ') plt.scatter (x[0:50],example_1, marker= '. ') Olor= ' B ', label= ' predict value ', s=10) Plt.scatter (x[0:50], y1[0][0:50], marker= '. ', color= ' g ', label= ' real value ', s= Plt.legend (loc= ' upper right ') plt.figure (2) plt.xlabel (' Wavelength ') plt.ylabel (' Flux ') Plt.scatte R (X[0:50], example_6, marker= '. ', color= ' B ', label= ' predict value ', s=10) Plt.scatter (x[0:50], y1[5][0:50], marker= '. ' , color= ' g ', label= ' real value ', s=10-plt.legend (loc= ' upper right ') plt.show () def main (): X,y=read1 () x 1,y1=read2 () example_1, example_6, Error1 = Error (x,y, y1) figure (x,example_1,example_6,y1) If __name__ = ' __mai N__ ': Main ()