آموزش Recurrent Neural Networks
Get started with recurrent neural network (RNN) concepts in a simplified way and build simple applications with RNNs and Keras. RNN is a fast-growing domain within the AI world. Popular groundbreaking applications like language translation, speech synthesis, question answering, and text generation use RNNs as their base technology. Studying this technology, however, has several challenges. Most learning resources are math heavy and are difficult to navigate without good math skills. IT professionals from varying backgrounds need a simplified resource to learn the concepts and build models quickly. In this course, Kumaran Ponnambalam provides a simplified path to studying the basics of recurrent neural networks, allowing you to become productive quickly. Kumaran starts with a simplified introduction of RNN before walking through the process of building a model. He then covers the popular building blocks of RNN with GRUs, LSTMs, word embeddings, and transformers.
Introduction
Getting started with RNNs
()
Scope and prerequisites for the course
()
Setting up exercise files
()
1. Introduction to RNNs
A review of deep learning
()
Why sequence models?
()
A recurrent neural network
()
Types of RNNs
()
Applications of RNNs
()
2. RNN Concepts
Training RNN models
()
Forward propagation with RNN
()
Computing RNN loss
()
Backward propagation with RNN
()
Predictions with RNN
()
3. An RNN Example
A simple RNN example: Predicting stock prices
()
Data preprocessing for RNN
()
Preparing time series data with lookback
()
Creating an RNN model
()
Testing and predictions with RNN
()
4. RNN Architectures
The vanishing gradient problem
()
The gated recurrent unit
()
Long short-term memory
()
Bidirectional RNNs
()
5. An LSTM Example
Forecasting service loads with LSTM
()
Time series patterns
()
Preparing time series data for LSTM
()
Creating an LSTM model
()
Testing the LSTM model
()
Forecasting service loads: Predictions
()
6. Word Embeddings
Text based models: Challenges
()
Intro to word embeddings
()
Pretrained word embeddings
()
Text preprocessing for RNN
()
Creating an embedding matrix
()
7. Spam Detection with Word Embeddings
Spam detection example for embeddings
()
Preparing spam data for training
()
Building the embedding matrix
()
Creating a spam classification model
()
Predicting spam with LSTM and word embeddings
()