4-Loss Reduction

4- Single neuron Weight Updation & Loss Reduction

Target and loss calculation

Linear Regression Model Fit (House Size vs Price)

x = 3
target = 16


w = 3
b = 1


prediction = w * x + b
loss1 = (target - prediction)**2
print("1st try weight:",w, "Prediction:", prediction , "target:",target, "loss1:",loss1)


w = w + 1
print("increased the weight by one")


prediction = w * x + b
loss2 = (target - prediction)**2
print("2nd try weight:",w,"Prediction:", prediction , "target:",target, "loss2:",loss2)


if (loss1>loss2):
  print('loss is reduced by',(loss1-loss2)*100/loss1,"%")

x = 3 target = 16 w = 3 b = 1 prediction = w * x + b loss1 = (target - prediction)**2 print("1st try weight:",w, "Prediction:", prediction , "target:",target, "loss1:",loss1) w = w + 1 print("increased the weight by one") prediction = w * x + b loss2 = (target - prediction)**2 print("2nd try weight:",w,"Prediction:", prediction , "target:",target, "loss2:",loss2) if (loss1>loss2): print('loss is reduced by',(loss1-loss2)*100/loss1,"%")

automatic updation of weights and bias

Automatic Updation of weight is done by simply knowing if there is a loss, and if there is any, move the weight by a small value in direction which reduces the loss. This small value is called Learning Rate

Linear Regression Model Fit (House Size vs Price)

x = 3
target = 10


w = 2
b = 1


prediction = w * x + b
loss1 = (target - prediction)**2
print("1st try weight:",w, "| Prediction:", prediction , "| target:",target, "| loss1:",loss1)


# change the weight by a small value (learning rate) in direction which reduces the loss. 
w = w + loss1 * 0.1


prediction = w * x + b
loss2 = (target - prediction)**2
print("2nd try weight:",w, "| Prediction:", prediction , "| target:",target, "| loss2:",loss2)


if (loss1>loss2):
  print('loss is reduced by',(loss1-loss2)*100/loss1,"%")

x = 3 target = 10 w = 2 b = 1 prediction = w * x + b loss1 = (target - prediction)**2 print("1st try weight:",w, "| Prediction:", prediction , "| target:",target, "| loss1:",loss1) # change the weight by a small value (learning rate) in direction which reduces the loss. w = w + loss1 * 0.1 prediction = w * x + b loss2 = (target - prediction)**2 print("2nd try weight:",w, "| Prediction:", prediction , "| target:",target, "| loss2:",loss2) if (loss1>loss2): print('loss is reduced by',(loss1-loss2)*100/loss1,"%")