loss

import random
import time
import math
#import matplotlib.pyplot as plt
from mpl_toolkits.mplot3d import Axes3D
from matplotlib import cm
from matplotlib.ticker import LinearLocator, FormatStrFormatter
import matplotlib.pyplot as plt
import numpy as np


def loss(w0,w1, data):
        dataSum=0.0
        for d in data:
            xd = d[0]
            yd = d[1]
            hwxd=float(w0)+float(w1*xd)
            tmp = pow((yd-float(hwxd)),2)
            dataSum+= tmp
        #return(math.sqrt(1.0/(2.0*len(data))*dataSum))
        return(1.0/(2.0*len(data))*dataSum)
        #return(rmse(w0,w1, data))

def rmse(w0,w1, data):
        w0= w0 *1000
        w1=0.3047073401501218
        dataSum=0.0
        for d in data:
            xd = d[0]*1000
            yd = d[1]*1000
            hwxd=float(w0)+float(w1*xd)
            tmp = pow((yd-float(hwxd)),2)
            dataSum+= tmp
        return(math.sqrt(1.0/(2.0*len(data))*dataSum))
        #return(1.0/(2.0*len(data))*dataSum)

def gradDesc(eta,maxit,data):

    # wi := wi - eta(deriv(w))

    w0 = random.random()
    w1 = random.random()
    w0 = 0
    w1=0

    # w0=-32.632
    # w1=0.304315
    #
    # nw0 = -32.632

    threshold = 0.1
    iterations =0
    lossfunc=loss(w0,w1,data)
    print('initial loss =', loss(w0,w1,data))


    # fig = plt.figure()
    # ax = fig.gca(projection='3d')

    # plt.ion()
    # plt.show()
#    graph.plot(x,y,'ro',markersize=5)

    xarr=[]
    yarr=[]
    lossf=[]
    w0arr=[]
    w1arr=[]
    while(iterations < maxit):
        iterations+=1;
        print('iter:',iterations)

        w0sum=0.0
        for d in data:
            xd = d[0]
            yd = d[1]
            hwxd=float(w0)+float(w1*xd)
            tmp = (float(hwxd)-yd)
            w0sum+= tmp
        w0sum = (1.0/len(data))*w0sum
        nw0=w0-(eta*w0sum)


        w1sum=0.0
        for d in data:
            xd = d[0]
            yd = d[1]
            hwxd=float(w0)+float(w1*xd)
            tmp = (float(hwxd)-yd)*xd
            w1sum+= tmp
        w1sum = (1.0/len(data))*w1sum
        nw1=w1-(eta*w1sum)

        w0=nw0
        w1=nw1

        prevlossfunc = lossfunc
        lossfunc=loss(w0,w1,data)
        w0arr.append(w0)
        w1arr.append(w1)
        lossf.append(lossfunc)
        # plt.scatter(w1,lossfunc)
        #
        # plt.draw()
        #time.sleep(0.05)

    print(iterations,w0,w1,lossfunc, rmse(w0,w1,data))



    xmin = 1000000000000
    xmax =0
    for d in data:
        xarr.append(d[0])
        yarr.append(d[1])
        if (d[0] < xmin):
            xmin=d[0]
        if (d[0] > xmax):
            xmax=d[0]

    fig=plt.figure()
    graph= fig.add_subplot(111)
    graph.plot(xarr,yarr,'ro',markersize=5)

    #(w0+(xarr[0]*w1))
    graph.plot([xmin,xmax],[(w0+(xmin*w1)),(w0+(xmax*w1))])

    fig2=plt.figure()
    graph2= fig2.add_subplot(111)
    graph2.plot(w0arr,lossf,'r-',label='w0')

    graph2.plot(w1arr,lossf,'g-',label='w1')
    graph2.legend(loc=0)
    plt.show()


    # for d in data:
    #     xarr.append(d[0])
    #     yarr.append(d[1])
    # plt.scatter(xarr,yarr)
    #

#    plt.plot([200,1000], [((w1*200)+w0),((w1*1000)+w0)] )