ABSTRACT
Deep learning models perform well when there is enough data available for training, but otherwise the performance deteriorates rapidly owing to the so-called data shortage problem. Recently, model-agnostic meta-learning (MAML) was proposed to alleviate this problem by embedding common prior knowledge from different tasks into the initial parameters of the target model. Data shortages are very common in regional influenza predictions, and MAML also often struggles with regional influenza forecasting, especially when region-specific knowledge, such as peak timing or intensity, varies. In this paper, we propose a novel MAML-based parameter adjustment scheme for influenza forecasting, called MARAPAS. The fundamental idea of our scheme is to adjust the initial parameters obtained from common knowledge to a target region by using adjustment variables. We experimentally show that MARAPAS outperforms other MAML-based methods, in terms of root mean square error and Pearson correlation coefficient. Particularly, this scheme improves the forecasting performance by up to 34 % compared with that of the state-of-the-art schemes. We also show the robust forecasting accuracy of our scheme and demonstrate its applicability by performing zero-shot COVID-19 forecasting. © 2022 The Author(s)
ABSTRACT
Deep learning models perform well when there is enough data available for training, but otherwise the performance deteriorates rapidly owing to the so-called data shortage problem. Recently, model-agnostic meta-learning (MAML) was proposed to alleviate this problem by embedding common prior knowledge from different tasks into the initial parameters of the target model. Data shortages are very common in regional influenza predictions, and MAML also often struggles with regional influenza forecasting, especially when region-specific knowledge, such as peak timing or intensity, varies. In this paper, we propose a novel MAML-based parameter adjustment scheme for influenza forecasting, called MARAPAS. The fundamental idea of our scheme is to adjust the initial parameters obtained from common knowledge to a target region by using adjustment variables. We experimentally show that MARAPAS outperforms other MAML-based methods, in terms of root mean square error and Pearson correlation coefficient. Particularly, this scheme improves the forecasting performance by up to 34 % compared with that of the state-of-the-art schemes. We also show the robust forecasting accuracy of our scheme and demonstrate its applicability by performing zero-shot COVID-19 forecasting.
ABSTRACT
BACKGROUND: Nonpharmaceutical interventions (NPIs) have been implemented to suppress transmission of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). Evidence indicates that NPIs against coronavirus disease 2019 (COVID-19) may also have effects on transmission of seasonal influenza. METHODS: In this study, we use an absolute humidity-driven susceptible-infectious-recovered-susceptible (SIRS) model to quantify the reduction of influenza incidence and transmission in the United States and US Department of Health and Human Services regions after implementation of NPIs in 2020. We investigate long-term effect of NPIs on influenza incidence by projecting influenza transmission at the national scale over the next 5 years, using the SIRS model. RESULTS: We estimate that incidence of influenza A/H1 and B, which circulated in early 2020, was reduced by more than 60% in the United States during the first 10 weeks following implementation of NPIs. The reduction of influenza transmission exhibits clear geographical variation. After the control measures are relaxed, potential accumulation of susceptibility to influenza infection may lead to a large outbreak, the scale of which may be affected by length of the intervention period and duration of immunity to influenza. CONCLUSIONS: Healthcare systems need to prepare for potential influenza patient surges and advocate vaccination and continued precautions.