Introduction:
Generative models are gaining a lot of popularity recently among data scientists, mainly because they facilitate the building of AI systems that consume raw data from a source and automatically builds an understanding of it. Unlike supervised learning methods, generative models do not require labelled data.
Generative adversarial networks (GANs) are a special class of generative models introduced by Ian Goodfellow in 2014. In GAN, a generator function learns to synthesize samples that best resemble some dataset, while a discriminator function learns to distinguish between samples drawn from the dataset and samples synthesized by the generator. GANs have emerged as a promising framework for unsupervised learning: GAN generators are able to produce images of unprecedented visual quality, while GAN discriminators learn features with rich semantics that lead to state-of-the-art semi-supervised learning.
Course Objectives:
- Generative model using Auto Encoder
- GAN Architecture
- GAN Implementation on MNIST dataset
- Style Transfer with GAN
- Image Generation with GAN
Duration
7 hours, 1 Day Course
Mode of Delivery
Classroom-based, Instructor-led Training
Course Outlines
- Overview of Generative Models
- What is Generative Models
- Application of Generative Models
- Types of Generative Models
- Online Generative Model Demo with Sketch-RNN
- Installing Google Colab
- DeepDream
- Recap on Convolutional Neural Networks (CNN)
- Recap on Transfer Learning
- What is DeepDream?
- DeepDream Applications
- DeepDream Implementation
- Neural Style Transfer
- What is Neural Style Transfer?
- Neural Style Transfer Applications
- Neural Style Transfer Implementation
- Variational Autoencoder (VAE)
- What is Autoencoder
- Variational Autoencoder (VAE)
- VAE Implementation
- Generative Adversarial Networks (GAN)
- What is GAN?
- GAN Applications
- Basic DCGAN Architecture
- DCGAN Implementation
- GAN Challenges and Tricks
- Text Generation
- Recap on Recurrent Neural Networks (RNN)
- Recap on Long Short Term Memory (LSTM)
- Char by Char Text Generation with LSTM