Deep learning is a family of representation learning methods that construct multilayer neural network models to automatically extract hierarchical features from data. This chapter first reviews the historical development of deep learning, contrasting shallow models such as support vector machines and boosting with deep architectures that learn multiple levels of abstraction. It then introduces the basic ideas of deep learning, including unsupervised pre training and supervised fine tuning, and explains representative methods such as autoencoders, sparse coding, restricted Boltzmann machines, and deep belief networks. A detailed section is devoted to convolutional neural networks, covering their structure, parameter sharing, receptive fields, and invariance properties. Finally, the chapter presents a practical example of handwritten digit recognition using a CNN based on the LeNet 5 architecture, illustrating the complete workflow from data preprocessing to training and recognition.