In this session, you’ll study some variants of the original RRN architecture
First, you will study a novel RRN architecture -the Bidirectional RNN which is capable of reading sequences in the ‘reverse order’ as well and has proven to boost performance significantly.
Then you will study two important cutting-edge variants of the RNN which have it possible to train large networks on real datasets. In the previous session, you saw the although RNNs are capable of solving a variety of sequence problems, their architecture itself is their biggest enemy. you learnt about the problems of exploding and vanishing gradients which occur during the training of RNNs. In this section, you’ll learn how this problem is solved by two popular gated RNN architectures – the long, Short Term Memory (LSTM) and the Gated Recurrent Unit (GRU).
In this session:
- Bidirectional RNNs
- LSTMs and GRUs
Prerequisites
There are no prerequisites for this session other than knowledge of the previous modules on Neural Networks and the courses Statistics and ML.
Guidelines for In-Module Questions
The in-video and in-content questions for this module are not graded. Note that graded questions are given on a separate page labelled ‘Graded Questions’ at the end of this session. The graded questions in this session will adhere to the following guidelines:
People you will hear from in this session:
Subject Matter Expert:
Professor, IIIT-Bangalore
The International Institute of Information Technology, Bangalore, commonly known as IIIT Bangalore, is a premier national graduate school in India. Founded in 1999, it offers Integrated M.Tech., M.Tech., M.S. (Research) and PhD programs in the field of information Technology.
Report an error