Skip to main content
Crowdfunding
Python + AI for Geeks
Practice

Creating a Simple Linear Regression Model with TensorFlow

Linear regression is a machine learning algorithm that models the linear relationship (a relationship that can be represented by a straight line) between an input and an output.

For example, consider the following relationship:

  • When x is 1, y is 2

  • When x is 2, y is 4

  • When x is 3, y is 6

This relationship can be expressed as a linear relationship like y = 2x.

The code below is an example of creating a simple machine learning model that helps a computer learn the pattern of numbers.

To break it down simply, the computer learns to output twice the number it is given.

Linear Regression Model Example
import tensorflow as tf
import numpy as np

# Define input data
x_train = np.array([1, 2, 3, 4, 5], dtype=np.float32)
# Define output data
y_train = np.array([2, 4, 6, 8, 10], dtype=np.float32)

# Create model
model = tf.keras.Sequential([
    # Create a Dense layer with 1 input value and 1 output value
    tf.keras.layers.Dense(units=1, input_shape=[1])
])

# Compile model
model.compile(optimizer='sgd', loss='mean_squared_error')

# Train the model by repeating the data 100 times to learn the rule
model.fit(x_train, y_train, epochs=100, verbose=1)

# Make a prediction
print(model.predict([6]))  # Expected output: A value close to 12

Let's go over each part of the code.


1. Defining Input and Output Data​

x_train is the input data (1, 2, 3, 4, 5), and y_train is the output data (2, 4, 6, 8, 10).

In other words, the output is twice the input value.


2. Creating the Model​

Use tf.keras.Sequential() to create a neural network model.

Dense(units=1, input_shape=[1]) means a layer with one input and one output.


3. Compiling the Model​

optimizer='sgd' indicates using Stochastic Gradient Descent (SGD) for training. SGD helps minimize the loss function.

loss='mean_squared_error' is a method for reducing error.


Training the Model​

model.fit(x_train, y_train, epochs=100, verbose=1) trains the model by repeating the data 100 times to learn the rule.


Making Predictions​

When you execute model.predict([6]), it predicts a value close to twice 6, which is 12.


We've now created a simple linear regression model.

In the next class, we'll go through a simple quiz to review what we've learned about TensorFlow.

Want to learn more?

Join CodeFriends Plus membership or enroll in a course to start your journey.