Computer Science & AI · 2014
Generative Adversarial Nets
Ian J. Goodfellow, et al.
Overview
Goodfellow and colleagues introduced generative adversarial networks (GANs), a way to train generative models by pitting two networks against each other — one creating fake data, the other trying to detect it — until the fakes are indistinguishable from real data.
Founded adversarial generative modelling, a precursor to today's generative AI.
Key findings
Methods
A minimax training procedure: a generator network maps random noise to samples while a discriminator network learns to distinguish real from generated data, the two optimised against each other by gradient descent.
Keywords
Related papers
Computer Science & AI
Highly Accurate Protein Structure Prediction with AlphaFold
AlphaFold solved a 50-year grand challenge in biology: predicting a protein's three-dimensional shape from its amino-acid sequence. Its accuracy approached that of experimental methods, transforming structural biology.
Computer Science & AI
Attention Is All You Need
The paper that introduced the Transformer, an architecture built entirely on attention mechanisms with no recurrence or convolution. It became the foundation of modern large language models such as GPT and BERT.
Computer Science & AI
Deep Learning
A landmark review by three pioneers of the field, synthesising how deep neural networks learn layered representations of data. It explains why depth works and surveys the advances that made deep learning dominant across vision, speech, and language.