Brain-inspired hyperdimensional computing (HDC) has been introduced as an alternative computing paradigm to achieve efficient and robust learning. HDC simulates cognitive tasks by mapping all data points to patterns of neural activity in the high-dimensional space, which has demonstrated promising performances in a wide range of applications such as robotics, biomedical signal processing, and genome sequencing. Language tasks, generally solved using machine learning methods, are widely deployed on low-power embedded devices. However, existing HDC solutions suffer from major challenges that impede the deployment of low-power embedded devices: the storage and computation overhead of HDC models grows dramatically with (i) the number of dimensions and (ii) the complex similarity metric during the inference.

In this paper, we proposed a novel ensemble framework for the language task, termed L3E-HD, which enables efficient HDC on low-power edge devices. L3E-HD accelerates the inference by mapping data points to a high-dimensional binary space to simplify similarity search, which dominates costly and frequent operation in HDC. Through marrying HDC with the ensemble technique, L3E-HD also addresses the severe accuracy degradation induced by the compression of the dimension and precision of the model. Our experiments show that the ensemble technique is naturally a perfect fit to boost HDCs. We find that our L3E-HD, which is faster, more efficient, and more accurate than conventional machine learning methods, can even surpass the accuracy of the full-precision model at a smaller model size. Code is released at: https://github.com/MXHX7199/SIGIR22-EnsembleHDC.