 SharedPublic Example.sagewsOpen in CoCalc
Authors: Tien Chih, Demitri Plessas
Views : 2
#We first create the raw data matrix listsize=5 # This determines the number of boxers being compared rawdata=matrix(QQ,listsize,listsize); #This creates a zero square matrix of the appropriate size, the entries of this matrix are allowed to be rational. #We then input the win/loss data into the matrix, the += command increments the appropriate entry by the given amount. If boxer i loses to boxer j, we encode that by rawdata[i,j]+=1. If Boxer i and j tie, we encode that by rawdata[i,j]+=1/2 and rawdata[j,i]+=1/2 #Recall that in Sage, arrays are enumerated starting at 0, so the first entry is the 0th entry, rather than the 1st. rawdata[0,1]+=1 rawdata[0,3]+=1 #These encode the losses suffered by A rawdata[1,2]+=1 rawdata[1,4]+=1/2 rawdata[1,4]+=1 #These encode the losses and tie by boxer B # Boxers C and D didnot suffer ties or losses and so there are no entries for those rows rawdata[4,1]+=1/2 #This encodes the tie by E to C #The following lines encodes an adjacency matrix adj for the underlying unweighted, undirected graph, should there be some interest in it's structure. adj=matrix(QQ,listsize,listsize); for i in range(listsize): for j in range (listsize): if rawdata[i,j]!=0: adj[i,j]=1 adj[j,i]=1 #The following lines encodes an adjacency matrix adj for the underlying unweighted, directed graph, should there be some interest in it's structure. dij=matrix(QQ,listsize,listsize); for i in range(listsize): for j in range (listsize): if rawdata[i,j]!=0: dij[i,j]=1 #The following lines creates the normalized data matrix, where multiple-fights are normalized normal=matrix(QQ,listsize,listsize) for i in range(listsize): for j in range (listsize): if rawdata[i,j]+rawdata[j,i]!=0: normal[i,j]=rawdata[i,j]/(rawdata[i,j]+rawdata[j,i]) #If you wish to the normal matrix, comment out the first line under this if listsize<20. Otherwise, comment out the following 2 lines #normal #Uncomment this line to see normal if listsize<20 #for i in range (listsize): #Uncomment this and the following line if listsize >=20. # print normal.rows()[i] #We then create the stochastic data matrix: stoch=matrix(QQ,listsize,listsize) for i in range(listsize): rowsum=0 for j in range(listsize): rowsum=rowsum+normal[i,j] if rowsum!=0: for k in range(listsize): stoch[i,k]=normal[i,k]/rowsum #This stochasticizes rows for non-undefeated fighters. else: for l in range(listsize): stoch[i,l]=1/listsize #This creates the rows for undefeated fighters and allows stoch to be a true markov matrix #If you wish to the stoch matrix, comment out the first line under this if listsize<20. Otherwise, comment out the following 2 lines #stoch #Uncomment this line to see normal if listsize<20 #for i in range (listsize): #Uncomment this and the following line if listsize >=20. # print stoch.rows()[i] #We then create the positive stochastic matrix M J=matrix(QQ,listsize,listsize); for i in range(listsize): for j in range(listsize): J[i,j]=1 #This creates a square matrix where each value is 1 p=15/100 #The damping factor p M=(1-p)*stoch+p*(1/listsize)*J #The final matrix M #If you wish to the M matrix, comment out the first line under this if listsize<20. Otherwise, comment out the following 2 lines #M #Uncomment this line to see normal if listsize<20 #for i in range (listsize): #Uncomment this and the following line if listsize >=20. # print M.rows()[i] #The following lines identifies which eigenvalue/vector pair has eigenvalue 1 sizespec=len(M.eigenvectors_left()) stable=0 for i in range(sizespec): if M.eigenvectors_left()[i]=1: stable=i #We then list the entries of the appropriate eigenvector in order, mutiplying by 1.0 gives a decimal approximation for easier comparison for i in range(len(M.eigenvectors_left()[stable])): M.eigenvectors_left()[best][i]*1.0 #We then display the digraph based on the normalized data matrix so as to visualize the win/loss relationship between boxers G = DiGraph(normal) H = G.plot(color_by_label=True) H.show()