Java Deep Learning Projects

Titanic survival revisited with DL4J

In the preceding chapter, we solved the Titanic survival prediction problem using Spark-based MLP. We also saw that by using Spark-based MLP, the user has very little transparency of using the layering structure. Moreover, it was not explicit to define hyperparameters and so on.

Therefore, what I have done is used the training dataset and then performed some preprocessing and feature engineering. Then I randomly split the pre-processed dataset into training and testing (to be precise, 70% for training and 30% for testing). First, we create the Spark session as follows:

SparkSession spark = SparkSession.builder()                  .master("local[*]")                  .config("spark.sql.warehouse.dir", "temp/")// change accordingly .appName("TitanicSurvivalPrediction") ...