Quaternion Recurrent Neural Networks

Abstract

Recurrent neural networks (RNNs) are powerful architectures to model sequential data, due to their capability to learn short and long-term dependencies between the basic elements of a sequence. Nonetheless, popular tasks such as speech or images recognition, involve multi-dimensional input features that are characterized by strong internal dependencies between the dimensions of the input vector. We propose a novel quaternion recurrent neural network (QRNN), alongside with a quaternion long-short term memory neural network (QLSTM), that take into account both the external relations and these internal structural dependencies with the quaternion algebra. Similarly to capsules, quaternions allow the QRNN to code internal dependencies by composing and processing multidimensional features as single entities, while the recurrent operation reveals correlations between the elements composing the sequence. We show that both QRNN and QLSTM achieve better performances than RNN and LSTM in a realistic application of automatic speech recognition. Finally, we show that QRNN and QLSTM reduce by a maximum factor of 3.3x the number of free parameters needed, compared to real-valued RNNs and LSTMs to reach better results, leading to a more compact representation of the relevant information.

arXiv commentICLR Update - Full rework

Cited by in corpus (22)

• SpeechBrain: A General-Purpose Speech Toolkit
• Quaternion Knowledge Graph Embeddings
• Quaternion Convolutional Neural Networks for Heterogeneous Image Processing
• A Theoretical Perspective on Hyperdimensional Computing
• Compacter: Efficient Low-Rank Hypercomplex Adapter Layers
• Quaternion Convolutional Neural Networks for Detection and Localization of 3D Sound Events
• Beyond Fully-Connected Layers with Quaternions: Parameterization of Hypercomplex Multiplications with $1/n$ Parameters
• Quaternion Generative Adversarial Networks
• Compressing deep quaternion neural networks with targeted regularization
• Lightweight and Efficient Neural Natural Language Processing with Quaternion Networks
• Deep Quaternion Features for Privacy Protection
• Analyzing Knowledge Graph Embedding Methods from a Multi-Embedding Interaction Perspective
• Quaternion Graph Neural Networks
• SurReal: Complex-Valued Learning as Principled Transformations on a Scaling and Rotation Manifold
• Flexible Transmitter Network
• Quaternion Collaborative Filtering for Recommendation
• Generalizing Complex/Hyper-complex Convolutions to Vector Map Convolutions
• Quaternion Neural Networks for Multi-channel Distant Speech Recognition
• A Quaternion-Valued Variational Autoencoder
• Rotation-Invariant Gait Identification with Quaternion Convolutional Neural Networks
• Quaternion-Based Graph Convolution Network for Recommendation
• Quaternion Factorization Machines: A Lightweight Solution to Intricate Feature Interaction Modelling