Advanced 28 hours 13 lessons
Advanced Neural Networks
Architectures behind today's most capable models
Go deep. Master the architectures — CNNs, transformers, attention — that power modern vision and language models, and learn to train them well.
Learning objectives
By the end of this course, you'll be able to:
- Design convolutional networks for vision tasks
- Explain self-attention and why transformers changed everything
- Diagnose and fix training instabilities
- Fine-tune a pretrained model for your own task
What you'll build
Project 01
Image Classifier with a CNN
Build and train a convolutional network, then visualise what each layer has learned to see.
Project 02
Mini Transformer for Text
Implement attention from scratch and train a small transformer to continue a sentence.
Syllabus
4 modules · 13 lessons
- 01
Seeing: Convolutional Networks
- Convolutions, filters and feature maps
- Pooling, padding and stride
- Modern CNN architectures
- 02
Attention & Transformers
- The attention mechanism
- Multi-head self-attention
- Positional encoding
- Anatomy of a transformer block
- 03
Training At Scale
- Normalisation and residual connections
- Learning-rate schedules
- Regularisation that actually helps
- 04
Transfer & Fine-Tuning
- Why pretraining works
- Fine-tuning strategies
- Capstone: adapt a pretrained model to a new task
Ready to start Advanced Neural Networks?
Jump in for free and learn at your own pace. Your progress is saved as you go.
Enroll Free