FTRLProx_Hashing: Feature Hashing for FTRL-Proximal Model

In yanyachen/rFTRLProximal: FTRL-Proximal Algorithm

Description

Feature Hashing for FTRL-Proximal Algorithm. A wrapper of hashed.model.matrix function in the FeatureHashing package. Please always use this function to generate sparse matrix for training and prediction.

Usage

FTRLProx_Hashing(formula = ~., data, hash.size, signed.hash, verbose = TRUE,
  label = NULL)

Arguments

formula	formula or a character vector of column names (will be expanded to a formula)
data	a data.frame or data.table. The original data.
hash.size	positive integer. The hash size of feature hashing.
signed.hash	logical value. Indicating if the hashed value is multipled by random sign. This will reduce the impact of collision. Disable it will enhance the speed.
verbose	logical value. Indicating if the progress bar is displayed or not.
label	a vector containing labels.

Value

constructed dataset, an object of class "ftrl.Dataset"

Examples

library(data.table)
library(FeatureHashing)
data(ipinyou)
m.train <- FTRLProx_Hashing(~ 0 + ., ipinyou.train[, -"IsClick", with = FALSE],
                            hash.size = 2^13, signed.hash = FALSE, verbose = TRUE,
                            label = as.numeric(ipinyou.train$IsClick))
m.test <- FTRLProx_Hashing(~ 0 + ., ipinyou.test[,-"IsClick", with = FALSE],
                           hash.size = 2^13, signed.hash = FALSE, verbose = TRUE,
                           label = as.numeric(ipinyou.test$IsClick))

yanyachen/rFTRLProximal documentation built on May 4, 2019, 2:30 p.m.