MP4 | Video: AVC 1280x720 | Audio: AAC 44KHz 2ch | Duration: 5 Hours 11M | 758 MB
Genre: eLearning | Language: English
Keras is an open source neural network library written in Python. It is capable of running on top of MXNet, Deeplearning4j, Tensorflow, CNTK, or Theano. Designed to enable fast experimentation with deep neural networks, it focuses on being minimal, modular, and extensible.
This course provides a comprehensive introduction to deep learning. We start by presenting some famous success stories and a brief recap of the most common concepts found in machine learning. Then, we introduce neural networks and the optimization techniques to train them. We’ll show you how to get ready with Keras API to start training deep learning models, both on CPU and on GPU. Then, we present two types of neural architecture: convolutional and recurrent neural networks
First, we present a well-known use case of deep learning: recommender systems, where we try to predict the "rating" or "preference" that a user would give to an item. Then, we introduce an interesting subject called style transfer. Deep learning has this ability to transform images based on a set of inputs, so we’ll morph an image with a style image to combine them into a very realistic result. In the third section, we present techniques to train on very small datasets. This comprises transfer learning, data augmentation, and hyperparameter search, to avoid overfitting and to preserve the generalization property of the network.
Finally, we complete this course by what Yann LeCun, Director at Facebook, considered as the biggest breakthrough in Machine Learning of the last decade: Generative Adversarial Networks. These networks are amazingly good at capturing the underlying distribution of a set of images to generate new images.