learner.deeplearning: Deep Learning with h2o tuned by mlr

In MarcoNiemann/kaggle_house: Analysing House Prices in Ames, Iowa and Build a Sales Price Prediction Model

Description

This method performs parameter tuning and feature selection on the provided data data set.

Usage

learner.deeplearning(data = data_train_numeric_clean_imputed)

Arguments

data	data.frame containing the data that should be used for deep learning. For optimal results an imputed and cleaned data set should be provided. The default data set is the KaggleHouse train data set after imputation and conversion into a numeric data.frame.

Details

This method trains an ANN based on the h2o package (more specificially the h2o.deeplearning) function. For the training it uses the given data and tries to adjust the hidden (number of hidden layers) and the rate (learning rate) parameters. On top feature selection will be performed to only keep those features actually contributing to the model. The final results will be saved to the learner.deeplearning_result.RData file. This way they can later be reused to extract the optimal parameters for a deeplearning ANN.

References

A. Candel, J. Lanford, E. LeDell, V. Parmar, A. Arora (2015). Deep Learning with H2O (Third Edit.) Publisher: H2O.ai, Inc. URL: http://h2o.gitbooks.io/deep-learning/

S. Aiello, T. Kraljevic and P. Maj (2016). h2o: R Interface for H2O URL: http://www.h2o.ai/

Examples

 KaggleHouse:::learner.deeplearning(data_train_numeric_clean_imputed)

MarcoNiemann/kaggle_house documentation built on May 7, 2019, 2:50 p.m.