R/s_LIHADBoost.R

In egenn/rtemis: Machine Learning and Visualization

# s_LIHADBoost.R
# ::rtemis::
# 2018 E.D. Gennatas www.lambdamd.org
# boosting learning.rate vs. hytree learning.rate
# ... added to allow "weights = NULL" from gridSearchLearn

#' Boosting of Linear Hard Additive Trees \[R\]
#'
#' Boost a Linear Hard Additive Tree (i.e. LIHAD, i.e. LINAD with hard splits)
#'
#' By default, early stopping works by checking training loss.
#'
# @inheritParams hytboost
#' @inheritParams s_GLM
#' @param learning.rate Float (0, 1] Learning rate for the additive steps
#' @param init Float: Initial value for prediction. Default = mean(y)
#' @param max.iter Integer: Maximum number of iterations (additive steps) to perform. Default = 10
#' @param trace Integer: If > 0, print diagnostic info to console
#' @param base.verbose Logical: `verbose` argument passed to learner
#' @param print.error.plot String or Integer: "final" plots a training and validation (if available) error curve at the
#' end of training. If integer, plot training and validation error curve every this many iterations
#' during training
#' @param print.base.plot Logical: Passed to `print.plot` argument of base learner, i.e. if TRUE, print error plot
#' for each base learner
#' @param ... Additional parameters to be passed to learner
#' 
#' @author E.D. Gennatas
#' @export


s_LIHADBoost <- function(x, y = NULL,
                       x.test = NULL, y.test = NULL,
                       # x.valid = NULL, y.valid = NULL,
                       resid = NULL,
                       boost.obj = NULL,
                       learning.rate = .5, # overwrite mod.params$learning.rate
                       case.p = 1,
                       # mod.params = setup.LIHAD(),
                       # ++ hytreew params ++
                       max.depth = 5,
                       gamma = .1,
                       alpha = 0,
                       lambda = 1,
                       lambda.seq = NULL,
                       minobsinnode = 2,
                       minobsinnode.lin = 10,
                       shrinkage = 1,
                       part.minsplit = 2,
                       part.xval = 0,
                       part.max.depth = 1,
                       part.cp = 0,
                       part.minbucket = 5,
                       # init = mean(y),
                       lin.type = c("glmnet", "cv.glmnet", "lm.ridge", "allSubsets", "forwardStepwise",
                                    "backwardStepwise", "glm", "sgd", "solve", "none"),
                       cv.glmnet.nfolds = 5,
                       which.cv.glmnet.lambda = "lambda.min",
                       # -- hytreew params --
                       # weights = NULL,
                       max.iter = 10,
                       tune.n.iter = TRUE,
                       # cv.n.iter = TRUE, # By default, CV to find best n.iter
                       earlystop.params = setup.earlystop(),
                       lookback = TRUE,
                       init = NULL,
                       .gs = FALSE,
                       grid.resample.params = setup.resample("kfold", 5),
                       gridsearch.type = "exhaustive",
                       metric = NULL,
                       maximize = NULL,
                       cxrcoef = FALSE,
                       print.progress.every = 5,
                       print.error.plot = "final",
                       x.name = NULL,
                       y.name = NULL,
                       question = NULL,
                       base.verbose = FALSE,
                       verbose = TRUE,
                       grid.verbose = FALSE,
                       trace = 0,
                       prefix = NULL,
                       plot.fitted = NULL,
                       plot.predicted = NULL,
                       plot.theme = rtTheme,
                       print.plot = FALSE,
                       print.base.plot = FALSE,
                       print.tune.plot = TRUE,
                       plot.type = 'l',
                       save.gridrun = FALSE,
                       outdir = NULL,
                       n.cores = rtCores,
                       save.mod = ifelse(!is.null(outdir), TRUE, FALSE), ...) {

  # Intro ----
  if (missing(x)) {
    print(args(boost))
    return(invisible(9))
  }
  if (!is.null(outdir)) outdir <- paste0(normalizePath(outdir, mustWork = FALSE), "/")
  logFile <- if (!is.null(outdir)) {
    paste0(outdir, "/", sys.calls()[[1]][[1]], ".", format(Sys.time(), "%Y%m%d.%H%M%S"), ".log")
  } else {
    NULL
  }
  start.time <- intro(verbose = verbose, logFile = logFile)
  mod.name <- "LIHADBoost"
  lin.type <- match.arg(lin.type)

  # Dependencies ----
  dependency_check("rpart", "glmnet")

  # Arguments ----
  if (is.null(x.name)) x.name <- getName(x, "x")
  if (is.null(y.name)) y.name <- getName(y, "y")
  if (!verbose) print.plot <- FALSE
  verbose <- verbose | !is.null(logFile)
  # if (save.mod & is.null(outdir)) outdir <- paste0("./s.", mod.name)
  if (!is.null(outdir)) outdir <- paste0(normalizePath(outdir, mustWork = FALSE), "/")
  # extra.args <- list(...)
  # mod.params <- c(mod.params, extra.args)
  # mod.params$max.depth <- max.depth
  # mod.params$learning.rate <- learning.rate
  # if (!is.null(force.n.iter)) max.iter <- force.n.iter

  # Data ----
  dt <- prepare_data(x, y,
                    x.test, y.test,
                    # x.valid = x.valid, y.valid = y.valid,
                    # ifw = ifw, ifw.type = ifw.type,
                    # upsample = upsample, resample.seed = resample.seed,
                    verbose = verbose)
  x <- dt$x
  y <- dt$y
  x.test <- dt$x.test
  y.test <- dt$y.test
  # x.valid <- dt$x.valid
  # y.valid <- dt$y.valid
  xnames <- dt$xnames
  type <- dt$type
  checkType(type, "Regression", mod.name)

  if (verbose) dataSummary(x, y, x.test, y.test, type)
  if (print.plot) {
    if (is.null(plot.fitted)) plot.fitted <- if (is.null(y.test)) TRUE else FALSE
    if (is.null(plot.predicted)) plot.predicted <- if (!is.null(y.test)) TRUE else FALSE
  } else {
    plot.fitted <- plot.predicted <- FALSE
  }
  if (is.null(init)) init <- mean(y)

  # Grid Search ----
  if (is.null(metric)) {
    if (type == "Classification") {
      metric <- "Balanced Accuracy"
      if (is.null(maximize)) maximize <- TRUE
    } else if (type == "Regression") {
      metric <- "MSE"
      if (is.null(maximize)) maximize <- FALSE
    }
  }

  if (is.null(maximize)) {
    maximize <- if (type == "Classification") TRUE else FALSE
  }

  # .final <- FALSE
  gc <- gridCheck(max.depth, learning.rate,
                  # hytreenow params
                  shrinkage,
                  alpha, lambda)
  if (!.gs && (gc || tune.n.iter )) {
    gs <- gridSearchLearn(x = x, y = y,
                          mod = mod.name,
                          resample.params = grid.resample.params,
                          grid.params = list(learning.rate = learning.rate,
                                             max.depth = max.depth,
                                             shrinkage = shrinkage,
                                             alpha = alpha,
                                             lambda = lambda),
                          fixed.params = list(max.iter = max.iter,
                                              earlystop.params = earlystop.params,
                                              lookback = lookback,
                                              lambda.seq = lambda.seq,
                                              minobsinnode = minobsinnode,
                                              minobsinnode.lin = minobsinnode.lin,
                                              part.minsplit = part.minsplit,
                                              part.xval = part.xval,
                                              part.max.depth = part.max.depth,
                                              part.cp = part.cp,
                                              part.minbucket = part.minbucket,
                                              lin.type = lin.type,
                                              cv.glmnet.nfolds = cv.glmnet.nfolds,
                                              which.cv.glmnet.lambda = which.cv.glmnet.lambda,
                                              .gs = TRUE),
                          search.type = gridsearch.type,
                          # weights = weights,
                          metric = metric,
                          maximize = maximize,
                          save.mod = save.gridrun,
                          verbose = verbose,
                          grid.verbose = grid.verbose,
                          n.cores = n.cores)

    max.depth <- gs$best.tune$max.depth
    learning.rate <- gs$best.tune$learning.rate
    max.iter <- gs$best.tune$n.steps

    # Now ready to train final full model
    # .final <- TRUE
    .gs <- FALSE
  } else {
    gs <- NULL
  }

  # LIHADBoost ----
  if (verbose) parameterSummary(init,
                                max.iter,
                                learning.rate,
                                newline.pre = TRUE)
                                # mod.params)
  if (trace > 0) msg2("Initial MSE =", mse(y, init))
  if (verbose) msg2("Training LIHADBoost...", newline.pre = TRUE)
  if (.gs) {
    .xval <- x.test # this is the validation set carved out of the training set during gridSearch
    .yval <- y.test
  } else {
    .xval <- .yval <- NULL
    # .xval <- x.valid # these may be null
    # .yval <- y.valid
  }
  mod <- hytboost(x = x, y = y,
                  x.valid = .xval, y.valid = .yval,
                  resid = resid,
                  boost.obj = boost.obj,
                  learning.rate = learning.rate,
                  case.p = case.p,
                  # mod.params = mod.params,
                  # ++ hytreew params ++
                  max.depth = max.depth,
                  gamma = gamma,
                  shrinkage = shrinkage,
                  alpha = alpha,
                  lambda = lambda,
                  lambda.seq = lambda.seq,
                  minobsinnode = minobsinnode,
                  minobsinnode.lin = minobsinnode.lin,
                  part.minsplit = part.minsplit,
                  part.xval = part.xval,
                  part.max.depth = part.max.depth,
                  part.cp = part.cp,
                  part.minbucket = part.minbucket,
                  lin.type = lin.type,
                  cv.glmnet.nfolds = cv.glmnet.nfolds,
                  which.cv.glmnet.lambda = which.cv.glmnet.lambda,
                  # -- hytreew params --
                  max.iter = max.iter,
                  earlystop.params = earlystop.params,
                  init = init,
                  cxrcoef = cxrcoef,
                  print.error.plot = print.error.plot,
                  print.progress.every = print.progress.every,
                  base.verbose = base.verbose,
                  verbose = verbose,
                  trace = trace,
                  prefix = prefix,
                  print.plot = print.plot,
                  plot.type = 'l')

  # if lookback, use best n.iter to get fitted and predicted
  if (.gs && lookback) {
    sni <- selectiter(mod$error.valid, mod$error, plot = print.tune.plot)
    n.iter <- sni$best.nsteps
    if (verbose) msg2("Selected", n.iter, "iterations based on smoothed",
                     ifelse(is.null(mod$error.valid), "training", "validation"), "loss curve")
    mod$selected.n.steps <- sni$best.nsteps
  } else {
    n.iter <- NULL # will use all iterations, will not be max.iter if earlystopping on training
  }

  # Fitted ----
  if (is.null(n.iter)) {
    fitted <- mod$fitted
  } else {
    fitted <- predict(mod, x, n.iter = n.iter)
  }
  error.train <- mod_error(y, fitted)
  if (verbose) errorSummary(error.train)

  # VALID ----
  # error.valid <- if (!is.null(y.valid)) mod$error.valid else NULL

  # Predicted ----
  predicted <- error.test <- NULL
  if (!is.null(x.test)) {
    if (verbose) cat("\n"); msg2("Getting predicted values...")
    predicted <- predict(mod, x.test, n.iter = n.iter)
    if (!is.null(y.test)) {
      error.test <- mod_error(y.test, predicted)
      if (verbose) errorSummary(error.test)
    }
  }

  # Outro ----
  parameters <- list(init = init,
                     max.iter = max.iter,
                     earlystop.params = earlystop.params,
                     learning.rate = learning.rate
                     # mod.params = mod.params
                     )
  extra <- list(gs = gs)
                # error.valid = error.valid)
  rt <- rtModSet(rtclass = "rtMod",
                 mod = mod,
                 mod.name = mod.name,
                 type = type,
                 parameters = parameters,
                 call = NULL,
                 y.train = y,
                 y.test = y.test,
                 x.name = x.name,
                 y.name = y.name,
                 xnames = xnames,
                 fitted = fitted,
                 se.fit = NULL,
                 error.train = error.train,
                 predicted = predicted,
                 se.prediction = NULL,
                 error.test = error.test,
                 varimp = NULL,
                 question = question,
                 extra = extra)

  rtMod.out(rt,
            print.plot,
            plot.fitted,
            plot.predicted,
            y.test,
            mod.name,
            outdir,
            save.mod,
            verbose,
            plot.theme)

  outro(start.time, verbose = verbose, sinkOff = ifelse(is.null(logFile), FALSE, TRUE))
  rt

} # rtemis::s_LIHADBoost