Research Article
Double-Layered Cortical Learning Algorithm for Time-Series Prediction
@INPROCEEDINGS{10.1007/978-3-030-92163-7_4, author={Takeru Aoki and Keiki Takadama and Hiroyuki Sato}, title={Double-Layered Cortical Learning Algorithm for Time-Series Prediction}, proceedings={Bio-Inspired Information and Communications Technologies. 13th EAI International Conference, BICT 2021, Virtual Event, September 1--2, 2021, Proceedings}, proceedings_a={BICT}, year={2022}, month={1}, keywords={Cortical learning algorithm Hierarchical temporal memory Time-series data prediction}, doi={10.1007/978-3-030-92163-7_4} }- Takeru Aoki
Keiki Takadama
Hiroyuki Sato
Year: 2022
Double-Layered Cortical Learning Algorithm for Time-Series Prediction
BICT
Springer
DOI: 10.1007/978-3-030-92163-7_4
Abstract
This work proposes a double-layered cortical learning algorithm. The cortical learning algorithm is a time-series prediction methodology inspired from the human neuro-cortex. The human neuro-cortex has a multi-layer structure, while the conventional cortical learning algorithm has a single layer structure. This work introduces a double-layered structure into the cortical learning algorithm. The first layer represents the input data and its context every time-step. The input data context presentation in the first layer is transferred to the second layer, and it is represented in the second layer as an abstract representation. Also, the abstract prediction in the second layer is reflected to the first layer to modify and enhance the prediction in the first layer. The experimental results show that the proposed double-layered cortical learning algorithm achieves higher prediction accuracy than the conventional single-layered cortical learning algorithms and the recurrent neural networks with the long short-term memory on several artificial time-series data.
