POLA: Online Time Series Prediction by Adaptive Learning Rates

TL;DR

POLA is an adaptive learning rate method for online time series prediction using recurrent neural networks, enabling models to quickly adapt to changing data patterns without overfitting.

Contribution

It introduces POLA, a meta-learning approach that automatically adjusts the learning rate of RNNs for online prediction in dynamic environments.

Findings

POLA achieves comparable or superior prediction accuracy.

POLA adapts effectively to changing data distributions.

POLA outperforms other online prediction methods.

Abstract

Online prediction for streaming time series data has practical use for many real-world applications where downstream decisions depend on accurate forecasts for the future. Deployment in dynamic environments requires models to adapt quickly to changing data distributions without overfitting. We propose POLA (Predicting Online by Learning rate Adaptation) to automatically regulate the learning rate of recurrent neural network models to adapt to changing time series patterns across time. POLA meta-learns the learning rate of the stochastic gradient descent (SGD) algorithm by assimilating the prequential or interleaved-test-then-train evaluation scheme for online prediction. We evaluate POLA on two real-world datasets across three commonly-used recurrent neural network models. POLA demonstrates overall comparable or better predictive performance over other online prediction methods.