Comparison of the Performance of Different Recurrent Neural Network Models on Sequence Classification: A Simulation Study

• Activity Number: 68 - Modern Statistical Learning Methods
• Type: Contributed
• Date/Time: Monday, August 3, 2020 : 10:00 AM to 2:00 PM
• Sponsor: Section on Statistical Learning and Data Science
• Author(s): Dawei Liu* and Charlie Cao
• Companies: Biogen and Biogen
• Keywords: recurrent neural network; lstm; gru; attention

Abstract

As an emerging machine learning paradigm, recurrent neural network (RNN) has become a powerful tool for analyzing sequence data. The long short-term memory (LSTM) and gated recurrent unit (GRU) are two popular RNN models for sequence classification. Recently, some new developments, such as attention, have been shown to improve the performance of LSTM and GRU. There are many different RNN models, but there is a shortage of systematic comparison of their performance. In this study, we will conduct a comprehensive simulation study in the context of sequence classification to evaluate the performance of LSTM and GRU, in the presence or absence of attention.