人工智能原理Lecture 8 生成性对抗网络 Generative Adversarial Networks -PPT精品课件.pptx

Lecture 8: Generative Adversarial Network 2 November 27, 2019 Artificial Intelligence Generative Adversarial Networks • Genarative • Learn a generative model • Adversarial • Trained in an adversarial setting • Networks • Use Deep Neural Networks Artificial Intelligence 3 Generative Models November 27, 2019 Artificial Intelligence 4 Generative Models November 27, 2019 Artificial Intelligence 5 Why Generative Models? • Discriminative models • Given a image X, predict a label Y • Estimates P(Y|X) • Discriminative models limitations: • Can’t model P(X) • Can’t generate new images • Generative models • Can model P(X) • Can generate new images November 27, 2019 Artificial Intelligence 6 Magic of GANs November 27, 2019 Artificial Intelligence 7 Magic of GANs • Which one is Computer generated? November 27, 2019 Artificial Intelligence 8 Magic of GANs November 27, 2019 Artificial Intelligence 9 GAN’s Architecture November 27, 2019 Artificial Intelligence 10 November 27, 2019 Adversarial Training • Adversarial Samples: • • • • We can generate adversarial samples to fool a discriminative model We can use those adversarial samples to make models robust We then require more effort to generate adversarial samples Repeat this and we get better discriminative model • GANs extend that idea to generative models: • • • • Generator: generate fake samples, tries to fool the Discriminator Discriminator: tries to distinguish between real and fake samples Train them against each other Repeat this and we get better Generator and Discriminator Artificial Intelligence 11 Training Discriminator November 27, 2019 Artificial Intelligence 12 Training Generator November 27, 2019 Artificial Intelligence 13 Mathematical formulation November 27, 2019 Artificial Intelligence 14 Mathematical formulation November 27, 2019 Artificial Intelligence 15 Mathematical formulation November 27, 2019 16 November 27, 2019 Artificial