Exercise

Imputing missing data

In this exercise, we will use GLRM to impute missing data. We are going to build a GLRM model from a dataset named fashion_mnist_miss, where 20% of values are missing. The goal is to fill these values by making a prediction using h2o.predict() with the GLRM model.

In this exercise an h2o instance is already running, so it is not necessary to call h2o.init().

The h2o package and fashion_mnist_miss have been loaded.

Instructions

100 XP

Store the input data in the h2o cluster.
Execute a rank-2 GLRM model on normalized input data and limit the number of iterations to 100.
Impute missing data by predicting with the previous model.
Inspect the basic statistics of the first five columns in the console.