logistic

import csv
import pandas as pd
import numpy as np
import itertools
from scipy.sparse import csr_matrix
from scipy.sparse import save_npz, load_npz
from pandas import read_csv
from scipy.sparse import hstack
from math import exp
sparse_mtx=csr_matrix(data_new)
import pandas as pd

data=pd.read_csv('training (2).csv',header=None)

#adding 1 to the data matrix for w0 and subracting 1 from class labels since range is from 0-19
A=np.ones(12000)
data.insert(0,column=0,  value=A)
data_class=data.iloc[:,data.shape[1]-1]
data_class=data_class-1

#dropping the 0 and last column from my data matrix 
data=data.drop(0 ,axis=1)
data=data.drop(61190 ,axis="columns")

#converting into sparse matrix for easy calculation and saving in into a file 
sparse_mtx=csr_matrix(data)
save_npz('training_sparse',sparse_mtx)

#initialize weights:K* n+1
W= np.zeros((20,61189))
W[:,:]=0.001


# prob matrix 
def probX(W,sparse_mtx):
    probarr=W@sparse_mtx.T
    probarr=probarr.expm1()

deltaarr=np.zeros((20,12000))
#delta matrix where deltarr[i][j]=1 for a class to which document belongs to (k*m)
    col=deltaarr.shape[1]
    for j in range (0,col):
        #print(j)
        c=data_class[j]
        #print(c)
        deltaarr[c][j]= 1.0
        #print(deltaarr[c][j])