The most common approach to imputing missing values in data tables involves analyzing either column-wise or row-wise relationships. These relationships are used to estimate missing entries based on available data from other columns in the same row or other rows in the same column. This paper introduces Ae \(^2\) I (Double Autoencoder for Imputation), a novel method that simultaneously and collaboratively leverages both row-wise and column-wise relationships to impute missing values. Empirical evaluations on the Movielens 1M dataset demonstrate that Ae \(^2\) I significantly outperforms existing state-of-the-art recommender system models.