Speaker : Anastasia Aidini (Postdoctoral Research)
Date : 16th of April
Location: FORTH
Paper Abstract:
We propose a novel method for imputing missing
data by adapting the well-known Generative Ad-
versarial Nets (GAN) framework. Accordingly,
we call our method Generative Adversarial Impu-
tation Nets (GAIN). The generator (G) observes
some components of a real data vector, imputes
the missing components conditioned on what is
actually observed, and outputs a completed vector.
The discriminator (D) then takes a completed vec-
tor and attempts to determine which components
were actually observed and which were imputed.
To ensure that D forces G to learn the desired
distribution, we provide D with some additional
information in the form of a hint vector. The hint
reveals to D partial information about the miss-
ingness of the original sample, which is used by
D to focus its attention on the imputation quality
of particular components. This hint ensures that G
does in fact learn to generate according to the true
data distribution. We tested our method on var-
ious datasets and found that GAIN significantly
outperforms state-of-the-art imputation methods.
Download Presentation