U21EC082 - Jinhal Maheshwari

In [1]: # Aim: To study and implement the simple linear regression with and without usin

In [2]: import numpy as np

import pandas as pd
import matplotlib.pyplot as plt

In [3]: data = pd.read_csv(r"C:\Users\aksha\Downloads\detaset.csv")

data

Out[3]: x y

0 0.00 -1.179169

1 0.01 0.515011

2 0.02 -0.047497

3 0.03 0.160731

4 0.04 0.704209

... ... ...

996 9.96 21.899947

997 9.97 21.150909

998 9.98 19.900154

999 9.99 22.883209

1000 10.00 19.949295

1001 rows × 2 columns

In [4]: x = data.iloc[:,0].values
y= data.iloc[:,-1].values

In [5]: x

Out[5]: array([ 0. , 0.01, 0.02, ..., 9.98, 9.99, 10. ])

In [6]: y

Out[6]: array([-1.17916881, 0.51501056, -0.04749717, ..., 19.90015398,

22.88320905, 19.94929548])

In [7]: plt.scatter(x,y)

Out[7]: <matplotlib.collections.PathCollection at 0x1ff82fc4810>

In [8]: from sklearn.model_selection import train_test_split

x_train,x_test,y_train,y_test = train_test_split(x,y,test_size = 0.25,random_sta

In [9]: w=2
b=1
mmse =0
y_pred = np.zeros(len(x))
for i in range(len(x)):
y_pred[i] = w*x[i]+b
mmse+=(y_pred[i]-y[i])**2

mmse = mmse/len(y_pred)
mmse
def compute_gradients(x, y, w, b):
m = len(y)
dw = (1/w) * np.sum((y_pred - y) * x)
db = (1/w) * np.sum(y_pred - y)
dw,db

In [10]: def train_linear_regression(x, y, learning_rate, epochs):

# Step 1: Initialize weights and bias
w = 0.0
b = 0.0

# Step 2: Calculate the number of samples

m = len(y)

# Step 3: Training loop

for epoch in range(epochs):
# Step 4: Calculate predictions
y_pred = w * x + b

# Step 5: Calculate loss (Mean Squared Error)

 mse = (1 / m) * sum((y - y_pred) ** 2)

# Step 6: Calculate gradients

dw = -(2 / m) * sum(x * (y - y_pred))
db = -(2 / m) * sum(y - y_pred)

# Step 7: Update weights and bias

w -= learning_rate * dw
b -= learning_rate * db

# Observe loss for each epoch

print(f"Epoch {epoch+1}/{epochs}, Loss: {mse:.4f}, w: {w:.4f}, b: {b:.4f

# Step 8: Print the final values of w and b

print(f"\nFinal values: w = {w:.4f}, b = {b:.4f}")

return w, b

In [11]: def predict(x, w, b):

return w * x + b

w, b = train_linear_regression(x_train, y_train, learning_rate=0.01, epochs=100)

# Predict values for the training and testing datasets
y_train_pred = predict(x_train, w, b)
y_test_pred = predict(x_test, w, b)

plt.figure(figsize=(10, 6))
# Plot training data and prediction line
plt.scatter(x_train, y_train, color='purple',s=10, label='Training Data')
plt.plot(x_train, y_train_pred, color='black', label='Training Fit Line')
# Plot test data and prediction line
plt.scatter(x_test, y_test, color='pink',s=10, label='Test Data')
plt.plot(x_test, y_test_pred, color='yellow', label='Test Fit Line')
# Plot settings
plt.title('Linear Regression on Training and Test Data')
plt.xlabel('x-axis')
plt.ylabel('y-axis')
plt.grid(True)
plt.legend()
plt.show()
Epoch 1/100, Loss: 140.4707, w: 1.3480, b: 0.2065
Epoch 2/100, Loss: 16.4519, w: 1.7978, b: 0.2762
Epoch 3/100, Loss: 2.6355, w: 1.9478, b: 0.3003
Epoch 4/100, Loss: 1.0962, w: 1.9977, b: 0.3091
Epoch 5/100, Loss: 0.9246, w: 2.0143, b: 0.3127
Epoch 6/100, Loss: 0.9054, w: 2.0197, b: 0.3147
Epoch 7/100, Loss: 0.9031, w: 2.0214, b: 0.3161
Epoch 8/100, Loss: 0.9027, w: 2.0218, b: 0.3173
Epoch 9/100, Loss: 0.9026, w: 2.0219, b: 0.3185
Epoch 10/100, Loss: 0.9024, w: 2.0218, b: 0.3196
Epoch 11/100, Loss: 0.9023, w: 2.0216, b: 0.3207
Epoch 12/100, Loss: 0.9022, w: 2.0215, b: 0.3218
Epoch 13/100, Loss: 0.9021, w: 2.0213, b: 0.3229
Epoch 14/100, Loss: 0.9019, w: 2.0212, b: 0.3240
Epoch 15/100, Loss: 0.9018, w: 2.0210, b: 0.3251
Epoch 16/100, Loss: 0.9017, w: 2.0208, b: 0.3262
Epoch 17/100, Loss: 0.9016, w: 2.0207, b: 0.3272
Epoch 18/100, Loss: 0.9015, w: 2.0205, b: 0.3283
Epoch 19/100, Loss: 0.9014, w: 2.0203, b: 0.3294
Epoch 20/100, Loss: 0.9012, w: 2.0202, b: 0.3304
Epoch 21/100, Loss: 0.9011, w: 2.0200, b: 0.3315
Epoch 22/100, Loss: 0.9010, w: 2.0199, b: 0.3325
Epoch 23/100, Loss: 0.9009, w: 2.0197, b: 0.3336
Epoch 24/100, Loss: 0.9008, w: 2.0195, b: 0.3346
Epoch 25/100, Loss: 0.9007, w: 2.0194, b: 0.3356
Epoch 26/100, Loss: 0.9006, w: 2.0192, b: 0.3366
Epoch 27/100, Loss: 0.9005, w: 2.0191, b: 0.3377
Epoch 28/100, Loss: 0.9004, w: 2.0189, b: 0.3387
Epoch 29/100, Loss: 0.9003, w: 2.0188, b: 0.3397
Epoch 30/100, Loss: 0.9002, w: 2.0186, b: 0.3407
Epoch 31/100, Loss: 0.9001, w: 2.0185, b: 0.3417
Epoch 32/100, Loss: 0.9000, w: 2.0183, b: 0.3427
Epoch 33/100, Loss: 0.8998, w: 2.0182, b: 0.3437
Epoch 34/100, Loss: 0.8997, w: 2.0180, b: 0.3446
Epoch 35/100, Loss: 0.8997, w: 2.0179, b: 0.3456
Epoch 36/100, Loss: 0.8996, w: 2.0177, b: 0.3466
Epoch 37/100, Loss: 0.8995, w: 2.0176, b: 0.3476
Epoch 38/100, Loss: 0.8994, w: 2.0174, b: 0.3485
Epoch 39/100, Loss: 0.8993, w: 2.0173, b: 0.3495
Epoch 40/100, Loss: 0.8992, w: 2.0171, b: 0.3505
Epoch 41/100, Loss: 0.8991, w: 2.0170, b: 0.3514
Epoch 42/100, Loss: 0.8990, w: 2.0168, b: 0.3523
Epoch 43/100, Loss: 0.8989, w: 2.0167, b: 0.3533
Epoch 44/100, Loss: 0.8988, w: 2.0165, b: 0.3542
Epoch 45/100, Loss: 0.8987, w: 2.0164, b: 0.3552
Epoch 46/100, Loss: 0.8986, w: 2.0163, b: 0.3561
Epoch 47/100, Loss: 0.8985, w: 2.0161, b: 0.3570
Epoch 48/100, Loss: 0.8985, w: 2.0160, b: 0.3579
Epoch 49/100, Loss: 0.8984, w: 2.0158, b: 0.3588
Epoch 50/100, Loss: 0.8983, w: 2.0157, b: 0.3597
Epoch 51/100, Loss: 0.8982, w: 2.0156, b: 0.3606
Epoch 52/100, Loss: 0.8981, w: 2.0154, b: 0.3615
Epoch 53/100, Loss: 0.8980, w: 2.0153, b: 0.3624
Epoch 54/100, Loss: 0.8980, w: 2.0152, b: 0.3633
Epoch 55/100, Loss: 0.8979, w: 2.0150, b: 0.3642
Epoch 56/100, Loss: 0.8978, w: 2.0149, b: 0.3651
Epoch 57/100, Loss: 0.8977, w: 2.0148, b: 0.3660
Epoch 58/100, Loss: 0.8976, w: 2.0146, b: 0.3668
Epoch 59/100, Loss: 0.8976, w: 2.0145, b: 0.3677
Epoch 60/100, Loss: 0.8975, w: 2.0144, b: 0.3686
Epoch 61/100, Loss: 0.8974, w: 2.0142, b: 0.3694
Epoch 62/100, Loss: 0.8973, w: 2.0141, b: 0.3703
Epoch 63/100, Loss: 0.8973, w: 2.0140, b: 0.3711
Epoch 64/100, Loss: 0.8972, w: 2.0138, b: 0.3720
Epoch 65/100, Loss: 0.8971, w: 2.0137, b: 0.3728
Epoch 66/100, Loss: 0.8970, w: 2.0136, b: 0.3737
Epoch 67/100, Loss: 0.8970, w: 2.0135, b: 0.3745
Epoch 68/100, Loss: 0.8969, w: 2.0133, b: 0.3753
Epoch 69/100, Loss: 0.8968, w: 2.0132, b: 0.3762
Epoch 70/100, Loss: 0.8967, w: 2.0131, b: 0.3770
Epoch 71/100, Loss: 0.8967, w: 2.0130, b: 0.3778
Epoch 72/100, Loss: 0.8966, w: 2.0128, b: 0.3786
Epoch 73/100, Loss: 0.8965, w: 2.0127, b: 0.3794
Epoch 74/100, Loss: 0.8965, w: 2.0126, b: 0.3802
Epoch 75/100, Loss: 0.8964, w: 2.0125, b: 0.3810
Epoch 76/100, Loss: 0.8963, w: 2.0123, b: 0.3818
Epoch 77/100, Loss: 0.8963, w: 2.0122, b: 0.3826
Epoch 78/100, Loss: 0.8962, w: 2.0121, b: 0.3834
Epoch 79/100, Loss: 0.8962, w: 2.0120, b: 0.3842
Epoch 80/100, Loss: 0.8961, w: 2.0119, b: 0.3850
Epoch 81/100, Loss: 0.8960, w: 2.0117, b: 0.3857
Epoch 82/100, Loss: 0.8960, w: 2.0116, b: 0.3865
Epoch 83/100, Loss: 0.8959, w: 2.0115, b: 0.3873
Epoch 84/100, Loss: 0.8958, w: 2.0114, b: 0.3881
Epoch 85/100, Loss: 0.8958, w: 2.0113, b: 0.3888
Epoch 86/100, Loss: 0.8957, w: 2.0112, b: 0.3896
Epoch 87/100, Loss: 0.8957, w: 2.0110, b: 0.3903
Epoch 88/100, Loss: 0.8956, w: 2.0109, b: 0.3911
Epoch 89/100, Loss: 0.8956, w: 2.0108, b: 0.3918
Epoch 90/100, Loss: 0.8955, w: 2.0107, b: 0.3926
Epoch 91/100, Loss: 0.8954, w: 2.0106, b: 0.3933
Epoch 92/100, Loss: 0.8954, w: 2.0105, b: 0.3940
Epoch 93/100, Loss: 0.8953, w: 2.0104, b: 0.3948
Epoch 94/100, Loss: 0.8953, w: 2.0103, b: 0.3955
Epoch 95/100, Loss: 0.8952, w: 2.0101, b: 0.3962
Epoch 96/100, Loss: 0.8952, w: 2.0100, b: 0.3969
Epoch 97/100, Loss: 0.8951, w: 2.0099, b: 0.3977
Epoch 98/100, Loss: 0.8951, w: 2.0098, b: 0.3984
Epoch 99/100, Loss: 0.8950, w: 2.0097, b: 0.3991
Epoch 100/100, Loss: 0.8950, w: 2.0096, b: 0.3998

Final values: w = 2.0096, b = 0.3998