Paper Information | SBC BOOKS

Fill in your paper information

Language (*)

Title (*)

Keywords (*)

Abstract (*)

Time series forecasting is moving to the edge, where models run on small IoT devices and must cope with noisy sensor streams. Hyperdimensional Computing (HDC) fits this setting, since it trains in a single pass using cheap, parallel operations. Current HDC forecasters, however, either use a simplistic delta rule or a linear Kalman Filter that assumes Gaussian noise and keeps a costly D × D covariance. We propose UKF-HD, the first formulation of the Unscented Kalman Filter inside HDC. The HDC prediction is linear in the weights, so the nonlinearity lies in the encoder; we therefore apply the unscented transform to the input window and push it through that encoder. The resulting gain accounts for the encoder nonlinearity and replaces the D × D covariance with a small p × p uncertainty over the input. We also propose GradHD, a gradient-based HDC forecaster, and its combination with the unscented front-end (GradHD+UKF). On multi-sensor IoT datasets corrupted by Gaussian, missing-value and Poisson noise, UKF-HD is more robust than the linear Kalman gain while using far less memory

Pages (*)

File Link

English Information

Title

Keywords

Abstract

(*) To change the order drag the item to the new position.

Authors

#	Name
1	Adriano de Souza Ferreira(adrianosf@id.uff.br)
2	Leandro Santiago de Araújo(leandro@ic.uff.br)

(*) To change the order drag the item to the new position.

Reference

#	Reference
1	Heddes, M., Nunes, I., Vergés, P., Kleyko, D., Abraham, D., Givargis, T., Nicolau, A., and Veidenbaum, A. (2022). Torchhd: An open source python library to support research on hyperdimensional computing and vector symbolic architectures
2	Hernández-Cano, A., Zhuo, C., Yin, X., and Imani, M. (2021). RegHD: Robust and efficient regression in hyper-dimensional learning system. In 2021 58th ACM/IEEE Design Automation Conference (DAC), pages 7–12
3	Julier, S. J. and Uhlmann, J. K. (2004). Unscented filtering and nonlinear estimation. Proceedings of the IEEE, 92(3):401–422
4	Kalman, R. E. (1960). A new approach to linear filtering and prediction problems. Journal of Basic Engineering, 82(1):35–45
5	Kanerva, P. (2009). Hyperdimensional computing: An introduction to computing in distributed representation with high-dimensional random vectors. Cognitive Computation, 1(2):139–159
6	Mejri, M., Amarnath, C., and Chatterjee, A. (2024). A novel hyperdimensional computing framework for online time series forecasting on the edge
7	Moreno, I. G., Yu, X., and Rosing, T. (2024). KalmanHD: Robust on-device time series forecasting with hyperdimensional computing. In 2024 29th Asia and South Pacific Design Automation Conference (ASP-DAC), pages 710–715
8	Wang, J. and Al Faruque, M. (2024). SMORE: Similarity-based hyperdimensional domain adaptation for multi-sensor time series classification. In Proceedings of the 61st ACM/IEEE Design Automation Conference (DAC ’24)

Paper Registration

Fill in your paper information

English Information

Authors

Reference