Introduction to Deep Neural Networks (DNNs)

About the Course

This course is aimed at anyone who has awareness of AI and machine learning and may even have introductory level knowledge of neural networks and their usage. It will introduce the fundamental concepts of Deep Neural Networks (DNNs) and the reasons behind different DNN architectures. You will learn about most common DNN architectures such as FeedForward Neural Networks (FNNs), Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs) and the Transformer architecture. The main emphasis of this course is on understanding different DNN architectures, their strengths and weaknesses and when to apply the most suitable architecture when designing an AI solution to a specific problem.

What Will Be Covered?

The one-day course will cover:

• Brief introduction to Neural Networks (NNs) and Deep Neural Networks (DNNs)
• Explanation of why different NN architectures exist
• Highlight likely effects of network width, depth and other architecture choices
• Introduction to common DNN architectures:
• FeedForward Neural Networks (FNNs)
• Convolutional Neural Networks (CNNs)
• Recurrent Neural Networks (RNNs)
• Transformers
• Explanation of which DNN architectures are best suited to which problem domains

Additional Information

Instructors: The course will be led by Dr Charaka Palansuriya, with colleagues from the Pilot-UKAIFA training team at EPCC.

Requirements: Attendees wishing to take part in the practical components of the course should bring a laptop. Demos and discussion will be available for those with no experience of Python or those who cannot bring a laptop.

Catering: Coffee and Tea will be available at breaks, but please note that this free course does not include lunch. There is a wide range of places to buy lunch near the training venue.

If you have any questions for this course, please contact Dr Charaka Palansuriya.