Large neural networks that learn hierarchical features.
Deep learning uses many-layer networks to learn features directly from raw data: CNNs for images, RNN/LSTM for sequences, Transformers for language and beyond, GANs and diffusion models for generation. Requires large data, GPUs and careful tuning; pretrained models allow transfer learning.
ResNet-50 classifies images with a deep convolutional network.
Deep learning uses many layers to learn features automatically, which has driven major advances in images, speech and language.
Large neural networks that learn hierarchical features.
Deep learning uses many layers to learn features automatically, which has driven major advances in images, speech and language.
It has many layers, so it can learn increasingly abstract features from raw data without manual feature design.
Millions of parameters must be tuned, and the model needs many examples and powerful hardware to train well.
Typical applications include detecting defects in microscope images with a convolutional network, transcribing speech from recordings and translating text between languages with large models.
Study "Overfitting and regularisation" first (topic 96 of 104 in Software & Computing). It covers the groundwork this topic builds on.
Embeddings. Batch normalisation. Fine-tuning.
ResNet-50 classifies images with a deep convolutional network.
The next topic is "Natural language processing and LLMs", which builds directly on this one.
It is classed as Advanced in Software & Computing (topic 97 of 104). It is an advanced topic. It assumes you are comfortable with the intermediate topics in Software & Computing.
Want to practise? Open the interactive course for quick checks, quizzes, simulations and progress tracking.