perceptron_learning.py

'''
# PERCEPTRON LEARNING BOLEAN AND , OR FUNCTIONS

A Perceptron is an Artificial Neuron. It is the simplest possible Neural Network. Neural Networks are the building blocks of Machine Learning.

![image](https://user-images.githubusercontent.com/80022302/223597791-3ef58982-9aca-4fde-9b0e-9da14bb88076.png)

Perceptron can learn only linearly seperable data.

Non linearly seperable data cannot be read by the perceptron.

The data is given in the program using input() statement.

The output obtained is:

![image](https://user-images.githubusercontent.com/80022302/223598263-9bc239d7-6dcd-46bb-8b6a-91c3f0f0fdd1.png)

'''


import numpy as np
import pandas as pd
from matplotlib import pyplot as plt

#Boolean AND
#dataset = pd.DataFrame({'input1':[-1,-1,1,1],'input2':[-1,1,-1,1],'result':[-1,-1,-1,1]})
#Boolean OR
dataset = pd.DataFrame({'input1':[-1,-1,1,1],'input2':[-1,1,-1,1],'result':[-1,1,1,1]})

print("Dataset is: ")
print(dataset)

#activation function
def activation_function(x):
  #if x>=0 1 is returned
  if x>=0:
    return 1
  #if x<0 -1 is returned
  else:
    return -1

 
#perceptron starts its learning with some random weights.

class Perceptron:
  #defining a constructor
  def __init__(self,n,actFunc):
    self.actFunc = actFunc
    self.bias=0
    self.n=n
  #predict function to predict the weights in each iteration
  def predict(self,ls,w):
    sum=0
    k=0
    #print("Inputs are : ")
    for elem in range(0,len(ls)-1):
      #calculate sum 
      sum += w[k]*ls[elem]
      k=k+1
    s=sum+self.bias  #adding bias initially
    a = self.actFunc(s)
    return a
  def learnBoolean(self,data,learningRate,n): 
    
    w=[]  #list to store weights
    print("Initail random weights are : ")  
    #choose random weights initillay as iterations goes on adjust the rando weights to obtain the original weights
    for i in range(0,n):
      k=np.random.random()
      w.insert(i,k)
      print(w[i])
      #print(i,"th random value is : ",w[i])
    #initialised with random weights
    count=0
    while(True):
      val=True
      res=-1 
      #storing the result val 
      y = data["result"] 
      #traversing over the list
      for i,j in data.iterrows():
        count=count+1
        res=res+1
       # print("res is: ",res) 
        ls = j.tolist()          #convertinh the input to list
        a =self.predict(ls,w)    #calling a predict function
        #print("predicted value is: " ,a)
        #print("actual value is: ",y[res])
        #if guessed output is not same as predicted then tune the weights as per learning rate
        if a != y[res]:
          val=False
          for i in range(0,n):
            delw = (y[res]-a)*learningRate*ls[i]    #tuning the weights
            w[i]=w[i]+delw                          #Adding the delta w to w
            self.bias=learningRate*(y[res]-a)       #changing the bias according to learning rate
            #print("The value when",ls[i],"is: ",w[i],"and i is: ",i)
        #print("------------------------------")
      if val==True:
         break
    #print(count)
    print("SO Final output is: ")       #obtaining the final output
    for i in range(0,self.n):
      print(w[i])
    return count


#creating object for perceptron class

learningRate = float(input("ENter the learning rate: "))  #taking learniing rate as input from user
p = Perceptron(2,activation_function)  #creating a perceptron object
learn = p.learnBoolean(dataset,learningRate,2)  #calling a learnbollean function
print("The number of iterations it took for perceptron to learn is :  ",learn) #total iterations printing