Linear regression: Example of a GLA that makes use of matrices

linear regression GLA.r

# include library
library(gtTranslator)
# define gla
gla <- MakeGLA(types = list(xType),
  constants = list(size = xType$size),
  representation = list(count = integer, XXT = translator::matrix(ncol = size, nrow = size), XY = xType),
  prototype = function(count = 0, XXT = zeros(size, size), XY = zeros(size)){},
  # Add item: get called for each tuple in independent processes
  AddItem = function(x, y) {
    count = count + 1L
    XXT = XXT + x * x$t()
    XY = XY + x * y
  },
  # Add state: aggregates results over the different processes
  AddState = function(o) {
    count = count + o$count
    XXT = XXT + o$XXT
    XY = XY + o$XY
  },
  # get results: executed once to obtain final result
  GetResult = function(result = JSON) {
    return(beta = XXT$i() * XY, count = count)
  })
# read lineitem data (TPCH data)  
data <- Read(lineitem10g)
# apply gla on data
agg <- gla(data, inputs = c(translator::MakeVector[](l_extendedprice, l_tax, l_discount), l_quantity))
# get result
result <- as.object(agg)