The ICA implementation by sklearn uses the same interface as PCA, so there is little to add. Note that there is no measure of explained variance because ICA does not compute components successively. Instead, each component aims to capture independent aspects of the data.