mb706 / autoxgboost3 Goto Github PK

View Code? Open in Web Editor NEW

4.0 4.0 0.0 383 KB

autoxgboost emulation for mlr3

License: Other

R 100.00%

autoxgboost3's People

Contributors

Stargazers

Watchers

autoxgboost3's Issues

entry for booster in param set missing

use task from mattermost file

axgb_settings = autoxgboost_space(task, tune.threshold = FALSE)
rsmp_inner = axgb_settings$resampling
learner = axgb_settings$learner
ps = axgb_settings$searchspace

# i added this and then it suddenly worked
ps$add(learner$param_set$params$xgboost.booster)
###########

ti = TuningInstance$new(task = task, learner = learner, resampling = rsmp_inner, param_set = ps, measures = msr("classif.ce"), terminator = term("evals", n_evals = 1))
tuner = tnr("random_search")
# here would be the error
tuner$tune(ti)

Bug for binary classification

wparam = ParamDbl$new("xgboosts.cale_pos_weight", -10, 10)

must be

wparam = ParamDbl$new("xgboost.scale_pos_weight", -10, 10)

I think the encode selector is wrong

Parameter xgboost.cale_pos_weight not available on dataset gisette

Code to reproduce the error (the first part is for downloading the data):

ll = mlr3misc::encapsulate("callr", function(task.id) {
  library(mlr)
  library(OpenML)

  setOMLConfig(arff.reader = "RWeka")
  OMLtask = convertOMLTaskToMlr(getOMLTask(task.id))
  data = getTaskData(OMLtask$mlr.task)
  rin = OMLtask$mlr.rin
  rdesc = OMLtask$mlr.rin$desc
  task.id = OMLtask$mlr.task$task.desc$id
  task.type = OMLtask$mlr.task$type
  target = OMLtask$mlr.task$task.desc$target

  return(list(task.id = task.id, task.type = task.type, data = data, rin = rin, rdesc = rdesc, target = target))
}
, .args = list(task.id = 167213))

res = ll$result
row.names(res$data) = as.integer(row.names(res$data))

if (res$task.type == "classif") {
  task = mlr3::TaskClassif$new(id = res$task.id, backend = res$data, target = res$target)
} else {
  task = mlr3::TaskRegr$new(id = res$task.id, backend = res$data, target = res$target)
}

library(autoxgboost)
library(mlr3tuning)
library(checkmate)
data="data/gisette"
task = readRDS(file.path(data, "task.rds"))
axgb_settings = autoxgboost_space(task, tune.threshold = FALSE)
rsmp_inner = axgb_settings$resampling
learner = axgb_settings$learner
ps = axgb_settings$searchspace
#ps$add(learner$param_set$params$xgboost.booster)
#ps$add(ParamFct$new(id = "xgboost.booster", levels = "gblinear", default = "gblinear"))
ti = TuningInstance$new(task = task, learner = learner, resampling = rsmp_inner, param_set = ps, measures = msr("classif.ce"), terminator = term("evals", n_evals = 1))
tuner = tnr("random_search")
tuner$tune(ti)

throwing the error:

> tuner$tune(ti)
INFO  [12:11:27.547] Starting to tune 9 parameters with '<TunerR
andomSearch>' and '<TerminatorEvals>'
INFO  [12:11:27.580] Terminator settings: n_evals=1
INFO  [12:11:27.632] Evaluating 1 configurations
INFO  [12:11:27.639]  xgboost.eta xgboost.gamma xgboost.max_dept
h
INFO  [12:11:27.639]    0.1015153     -6.913219                1
4
INFO  [12:11:27.639]  xgboost.colsample_bytree xgboost.colsample
_bylevel
INFO  [12:11:27.639]                 0.6309831                 0
.6852687
INFO  [12:11:27.639]  xgboost.lambda xgboost.alpha xgboost.subsa
mple
INFO  [12:11:27.639]       -7.631537      4.986284         0.741
8983
INFO  [12:11:27.639]  xgboosts.cale_pos_weight
INFO  [12:11:27.639]                 -4.246136
Error in (function (xs)  :
  Assertion on 'xs' failed: Parameter 'xgboosts.cale_pos_weight'
 not available..

using trafo forces breaking asserts

When adding a trafo to the parameter set, somehow undesired asserts make unnecessary checks which lead to an assert error.
This makes log sampling currently not possible for random search.
Steps to reproduce:

use same dataset as usual
execute the following code

library(autoxgboost)
library(mlr3tuning)
library(checkmate)
data="data/gisette"
task = readRDS(file.path(data, "task.rds"))
axgb_settings = autoxgboost_space(task, tune.threshold = FALSE)
rsmp_inner = axgb_settings$resampling
learner = axgb_settings$learner
ps = axgb_settings$searchspace
ps$add(
  ParamInt$new(
    id = "xgboost.nrounds",
    lower = as.integer(log(10, 3)),
    upper = as.integer(log(2430, 3)),
    tags = "budget"
  )
)
ps$trafo = function(x, param_set) {
  x$xgboost.nrounds = 3^x$xgboost.nrounds
  return(x)
}
ti = TuningInstance$new(task = task, learner = learner, resampling = rsmp_inner, param_set = ps, measures = msr("classif.ce"), terminator = term("evals", n_evals = 1))
tuner = tnr("random_search")
tuner$tune(ti)

While removing the trafo gives us no assert error:

library(autoxgboost)
library(mlr3tuning)
library(checkmate)
data="data/gisette"
task = readRDS(file.path(data, "task.rds"))
axgb_settings = autoxgboost_space(task, tune.threshold = FALSE)
rsmp_inner = axgb_settings$resampling
learner = axgb_settings$learner
ps = axgb_settings$searchspace
ps$add(
  ParamInt$new(
    id = "xgboost.nrounds",
    lower = as.integer(log(10, 3)),
    upper = as.integer(log(2430, 3)),
    tags = "budget"
  )
)
ti = TuningInstance$new(task = task, learner = learner, resampling = rsmp_inner, param_set = ps, measures = msr("classif.ce"), terminator = term("evals", n_evals = 1))
tuner = tnr("random_search")
tuner$tune(ti)

mb706 / autoxgboost3 Goto Github PK

autoxgboost3's People

Contributors

Stargazers

Watchers

autoxgboost3's Issues

entry for booster in param set missing

Bug for binary classification

I think the encode selector is wrong

Parameter xgboost.cale_pos_weight not available on dataset gisette

using trafo forces breaking asserts

Recommend Projects

React

Vue.js

Typescript

TensorFlow

Django

Laravel

D3

Recommend Topics

javascript

web

server

Machine learning

Visualization

Game

Recommend Org

Facebook

Microsoft

Google

Alibaba

D3

Tencent