Modifying memories in a Recurrent Neural Network Unit

Vlad Velici, Adam Prügel-Bennett

Abstract: Long Short-Term Memory (LSTM) units have the ability to memorise and use long-term dependencies between inputs to generate predictions on time series data. We introduce the concept of modifying the cell state (memory) of LSTMs using rotation matrices parametrised by a new set of trainable weights. This addition shows significant increases of performance on some of the tasks from the bAbI dataset.

TL;DR: Adding a new set of weights to the LSTM that rotate the cell memory improves performance on some bAbI tasks.

Keywords: LSTM, RNN, rotation matrix, long-term memory, natural language processing

Data: [bAbI](https://paperswithcode.com/dataset/babi-1)