This project demonstrates building and training a neural network to classify images from the MNIST dataset using TensorFlow and Keras.
The MNIST dataset consists of 70,000 grayscale images of handwritten digits (0-9). It is divided into:
Training set: 60,000 images Test set: 10,000 images
Input Layer: Flatten 28x28 images into a 1D array of 784 elements Hidden Layer: Dense layer with 128 neurons and ReLU activation Dropout Layer: Dropout rate of 0.2 Output Layer: Dense layer with 10 neurons (one for each digit)
The model is trained using the Adam optimizer and sparse categorical crossentropy loss function for 5 epochs.
The model's accuracy is evaluated on the test set.