Table 1 Methodologies of imputation methods. The table lists the modeling strategies and input features used by each of the models, as reported by the teams. The models include k-nearest neighbors (KNN), deep tensor factorization (DTF), autoencoders (AE), convolutional neural networks (CNN), hidden Markov models (HMM), and gradient-boosted decision trees (GBT). The authors of Aug2019Impute and CostaLab v2 did not describe their methods
Name | Model | Norm | Inputs | |||
|---|---|---|---|---|---|---|
Sequence | Functional | Average | Avocado | |||
Aug2019Impute | ||||||
BrokenNodes/v2 | KNN | arcsinh | \(\checkmark\) | |||
BrokenNodes v3 | KNN | arcsinh | \(\checkmark\) | \(\checkmark\) | ||
CostaLab v2 | ||||||
CUImpute1/CUWA/ICU | ensemble | arcsinh | \(\checkmark\) | \(\checkmark\) | \(\checkmark\) | |
Guacamole/Lavawizard | DTF | arcsinh | \(\checkmark\) | \(\checkmark\) | ||
HLYG/v1/v2 | GBT | quantile | \(\checkmark\) | \(\checkmark\) | \(\checkmark\) | |
imp/imp1 | DTF+AE | Cauchy | \(\checkmark\) | |||
KKT-ENCODE | CNN | arcsinh | \(\checkmark\) | |||
LiPingChun | DTF | arcsinh | \(\checkmark\) | \(\checkmark\) | ||
NittanyLions | KNN | \(\checkmark\) | ||||
NittanyLions2 | KNN | quantile | \(\checkmark\) | |||
SongLab | CNN | log1p | \(\checkmark\) | |||
SongLab2 | HMM | \(\checkmark\) | ||||
SongLab3 | CNN | log1p | \(\checkmark\) | \(\checkmark\) | \(\checkmark\) | |
UIOWA | CNN | quantile | \(\checkmark\) | \(\checkmark\) | ||