Deep Learning and Neural Networks

About this course

Deep learning is the technology behind today's most impressive AI: systems that recognise images, understand speech, translate languages and generate text. In this course you'll move from using AI to understanding and building it.

You'll start with how a single artificial neuron works and how networks learn through gradient descent and backpropagation, explained with intuition rather than heavy maths. Then you'll build networks in Keras: a digit recogniser, a convolutional network for images, and finally a crop disease classifier trained on real bean leaf photos collected in Uganda, using transfer learning from a model pretrained on millions of images.

You'll also learn how transformers, the architecture behind ChatGPT, process language, and use pretrained models for sentiment analysis and English-to-Swahili translation. Everything runs free on Google Colab GPUs.

By the end you will be able to

• Explain how neurons, layers, weights and activation functions work
• Describe how networks learn with loss functions, gradient descent and backpropagation
• Build, train and evaluate neural networks in Keras
• Use convolutional neural networks for image classification
• Prevent overfitting with validation, dropout, augmentation and early stopping
• Explain how transformers and attention process language
• Use pretrained models and transfer learning to solve real problems with limited data

Before you start

Python for AI and Machine Learning Essentials, or equivalent experience.