Skip to main content
Crowdfunding
Python + AI for Geeks
Practice

Training and Evaluating Models with Keras

In this lesson, we will explore how to train and evaluate a simple neural network model using Keras.


Data Preparation​

Keras provides a variety of built-in datasets that can be used to practice training deep learning models.

The example below demonstrates the process of training a neural network model using the handwritten digits dataset (MNIST).


Loading MNIST Data
from tensorflow.keras.datasets import mnist
from tensorflow.keras.utils import to_categorical

# Loading MNIST data
(x_train, y_train), (x_test, y_test) = mnist.load_data()

# Data preprocessing: normalization and one-hot encoding
x_train = x_train / 255.0
x_test = x_test / 255.0

# Reshape images from 28x28 (2D arrays) to 784-dimensional vectors (adjust for neural network input)
x_train = x_train.reshape(-1, 28 * 28)
x_test = x_test.reshape(-1, 28 * 28)

# Convert number labels (answers) to one-hot encoded format (e.g., number 3 → [0,0,0,1,0,0,0,0,0,0])
y_train = to_categorical(y_train, num_classes=10)
y_test = to_categorical(y_test, num_classes=10)

Code Explanation​

  1. Load the handwritten digits dataset using mnist.load_data().

  2. x_train and x_test are arrays of training and test images, respectively.

  3. y_train and y_test are the labels (answers) for the training and test data, respectively.

  4. Normalize the pixel values of x_train and x_test to a range of 0 to 1, which enhances the effectiveness of neural network training.

  5. Reshape the images in x_train and x_test from 28x28 2D arrays to 784-dimensional vectors to match the neural network input format.

  6. Use the to_categorical() function to convert labels to their one-hot encoded format. For example, the number 3 is converted to [0,0,0,1,0,0,0,0,0,0].

  7. num_classes=10 specifies the number of classes for the digits 0 through 9.


Model Training​

Now we will train the neural network model we previously defined.

Training the Model
# Train the model
model.fit(x_train, y_train, epochs=10, batch_size=32, validation_split=0.2)

Key Parameters​

  • epochs=10: Sets the number of epochs to 10, meaning the complete training data will be used 10 times for training.

  • batch_size=32: Updates weights using 32 samples per iteration.

  • validation_split=0.2: Uses 20% of the training data for validation.


Model Evaluation and Prediction​

Once training is complete, the trained model can be evaluated using the test data, and predictions can be made on new input data.

Evaluating and Predicting with the Model
# Evaluate the model
test_loss, test_acc = model.evaluate(x_test, y_test)

# Make predictions
predictions = model.predict(x_test)

Code Explanation​

  1. Use model.evaluate(x_test, y_test) to evaluate the model's performance on the test data.

  2. Call model.predict(x_test[:5]) to output predictions for some test data samples.


We have now learned the basic process of creating, training, and evaluating a neural network model using Keras.

In the next lesson, we'll solve a simple quiz based on what we've learned about Keras so far.

Want to learn more?

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