Lag Selection

Lag selection, the process of determining the optimal number of past data points to use in predictive models, is crucial for accurate forecasting across diverse fields. Current research emphasizes finding robust methods for selecting this parameter, particularly within the context of high-dimensional time series and deep learning architectures, focusing on techniques like cross-validation and convex programming to improve model performance. The accurate selection of lags significantly impacts forecasting accuracy, with both under- and overestimation leading to suboptimal results, highlighting the need for improved methodologies and evaluation metrics, such as length-adaptive measures for applications like simultaneous speech translation. This research directly impacts the reliability and efficiency of predictive models across various domains.

Papers

May 18, 2024

Lag Selection for Univariate Time Series Forecasting using Deep Learning: An Empirical Study
José Leites, Vitor Cerqueira, Carlos Soares
Deep Learning Empirical Study Time Series Forecasting Univariate Time Series Forecasting Superior Forecasting Lag Selection

March 3, 2023

Lag selection and estimation of stable parameters for multiple autoregressive processes through convex programming
Somnath Chakraborty, Johannes Lederer, Rainer von Sachs
Estimation Task Autoregressive Model High Dimensional Time Series Convex Program Square Root Lag Selection

June 12, 2022

Over-Generation Cannot Be Rewarded: Length-Adaptive Average Lagging for Simultaneous Speech Translation
Sara Papi, Marco Gaido, Matteo Negri, Marco Turchi
Simultaneous Speech Translation Multiple Generation Lag Selection

Lag Selection

Papers

Lag Selection for Univariate Time Series Forecasting using Deep Learning: An Empirical Study

Lag selection and estimation of stable parameters for multiple autoregressive processes through convex programming

Over-Generation Cannot Be Rewarded: Length-Adaptive Average Lagging for Simultaneous Speech Translation