diff --git a/DESCRIPTION b/DESCRIPTION
index 85253382c..a4309bd20 100644
--- a/DESCRIPTION
+++ b/DESCRIPTION
@@ -98,6 +98,7 @@ Roxygen: list(markdown = TRUE, r6 = FALSE)
 RoxygenNote: 7.3.2
 VignetteBuilder: knitr
 Collate:
+    'FilterEnsemble.R'
     'Graph.R'
     'GraphLearner.R'
     'mlr_pipeops.R'
diff --git a/R/FilterEnsemble.R b/R/FilterEnsemble.R
new file mode 100644
index 000000000..9a9076790
--- /dev/null
+++ b/R/FilterEnsemble.R
@@ -0,0 +1,105 @@
+
+
+#' @title Filter Ensemble
+#'
+#' @usage NULL
+#' @name mlr_filters_ensemble
+#' @format [`R6Class`][R6::R6Class] object inheriting from [`Filter`][mlr3filters::Filter].
+#'
+#' @description
+#' Implements the filte rensemble proposed in `r cite_bib("binder_2020")`.
+
+FilterEnsemble = R6Class("FilterEnsemble", inherit = mlr3filters::Filter,
+  public = list(
+    initialize = function(filters) {
+      private$.wrapped = assert_list(filters, types = "Filter", min.len = 1)
+      fnames = map_chr(private$.wrapped, "id")
+      names(private$.wrapped) = fnames
+      types_list = map(discard(private$.wrapped, function(x) test_scalar_na(x$task_types)), "task_types")
+      if (length(types_list)) {
+        task_types = Reduce(intersect, types_list)
+      } else {
+        task_types = NA_character_
+      }
+      .own_param_set = ps(
+        weights = p_uty(custom_check = crate(function(x) {
+          check_numeric(x, len = length(fnames)) %check&&%
+            (check_names(names(x), type = "unnamed") %check||%
+              check_names(names(x), type = "unique", permutation.of = fnames))
+          }, fnames),
+          tags = "required"
+        ),
+        rank_transform = p_lgl(init = FALSE, tags = "required")
+      )
+
+      super$initialize(
+        id = paste(fnames, collapse = "."),
+        task_types = task_types,
+        task_properties = unique(unlist(map(private$.wrapped, "task_properties"))),
+        param_set = ParamSetCollection$new(c(list(.own_param_set), map(private$.wrapped, "param_set"))),
+        feature_types = Reduce(intersect, map(private$.wrapped, "feature_types")),
+        packages = unique(unlist(map(private$.wrapped, "packages"))),
+        label = "meta",
+        man = "mlr3pipelines::mlr_filters_ensemble"
+      )
+    },
+    get_weights_tunetoken = function(normalize_weights = "uniform") {
+      assert_choice(normalize_weights, c("uniform", "naive", "no"))
+      paradox::to_tune(self$get_weights_search_space(normalize_weights = normalize_weights))
+    },
+    set_weights_to_tune = function(normalize_weights = "uniform") {
+      assert_choice(normalize_weights, c("uniform", "naive", "no"))
+      self$param_set$set_values(.values = list(weights = self$get_weights_tunetoken(normalize_weights = normalize_weights)))
+      invisible(self)
+    },
+    get_weights_search_space = function(weights_param_name = "weights", normalize_weights = "uniform", prefix = "w") {
+      assert_string(prefix)
+      assert_string(weights_param_name)
+      assert_choice(normalize_weights, c("uniform", "naive", "no"))
+      fnames = names(private$.wrapped)
+      innames = if (prefix == "") fnames else paste0(prefix, ".", fnames)
+      domains = rep(list(p_dbl(0, 1)), length(fnames))
+      names(domains) = innames
+
+      domains$.extra_trafo = crate(function(x) {
+        w = unlist(x[innames], use.names = FALSE)
+        names(w) = fnames
+        x[innames] = NULL
+
+        if (normalize_weights == "uniform") {
+          w[w > 1 - .Machine$double.eps] = 1 - .Machine$double.eps
+          w = -log1p(-w)
+          w = w / max(sum(w), .Machine$double.eps)
+        } else if (normalize_weights == "naive") {
+          w = w / max(sum(w), .Machine$double.eps)
+        }
+        x[[weights_param_name]] = w
+        x
+      }, innames, fnames, normalize_weights, weights_param_name)
+
+      do.call(paradox::ps, domains)
+    }
+  ),
+  private = list(
+    .wrapped = NULL,
+    .own_param_set = NULL,
+    .calculate = function(task, nfeat) {
+      pv = private$.own_param_set$get_values()
+      fn = task$feature_names
+      nfeat = length(fn)  # need to rank all features in an ensemble
+      weights = pv$weights
+      wnames = names(private$.wrapped)
+      if (!is.null(names(weights))) {
+        weights = weights[wnames]
+      }
+      scores = pmap(list(private$.wrapped, weights), function(x, w) {
+        x$calculate(task, nfeat)
+        s = x$scores[fn]
+        if (pv$rank_transform) s = rank(s)
+        s * w
+      })
+      structure(rowSums(as.data.frame(scores)), names = fn)
+    }
+  )
+
+)
diff --git a/R/bibentries.R b/R/bibentries.R
index de55741d5..0af04346f 100644
--- a/R/bibentries.R
+++ b/R/bibentries.R
@@ -52,5 +52,16 @@ bibentries = c(
     author    = "Yujun Wu and Dennis D Boos and Leonard A Stefanski",
     title     = "Controlling Variable Selection by the Addition of Pseudovariables",
     journal   = "Journal of the American Statistical Association"
+  ),
+
+  binder_2020 = bibentry("inproceedings",
+    doi       = "10.1145/3377930.3389815",
+    year      = "2020",
+    publisher = "Association for Computing Machinery",
+    pages     = "471--479",
+    author    = "Martin Binder and Julia Moosbauer and Janek Thomas and Bernd Bischl",
+    title     = "Multi-objective hyperparameter tuning and feature selection using filter ensembles",
+    booktitle = "Proceedings of the 2020 Genetic and Evolutionary Computation Conference"
   )
+
 )
diff --git a/R/zzz.R b/R/zzz.R
index a4333c8a8..a61b149b7 100644
--- a/R/zzz.R
+++ b/R/zzz.R
@@ -19,11 +19,24 @@ register_mlr3 = function() {
   x$pipeops$properties = c("validation", "internal_tuning")
 }
 
+register_mlr3filters = function() {
+  if ("mlr3filters" %in% loadedNamespaces()) {
+    x = utils::getFromNamespace("mlr_filters", ns = "mlr3filters")
+    mlr_filters$add("ensemble", FilterEnsemble)
+  }
+}
+
+
+
 paradox_info <- list2env(list(is_old = FALSE), parent = emptyenv())
 
 .onLoad = function(libname, pkgname) {  # nocov start
   register_mlr3()
-  setHook(packageEvent("mlr3", "onLoad"), function(...) register_mlr3(), action = "append")
+  register_mlr3filters()
+  setHook(packageEvent("mlr3", "onLoad"), function(...) {
+    register_mlr3()
+    register_mlr3filters()
+  }, action = "append")
   backports::import(pkgname)
 
   assign("lg", lgr::get_logger("mlr3/mlr3pipelines"), envir = parent.env(environment()))
diff --git a/man/mlr_filters_ensemble.Rd b/man/mlr_filters_ensemble.Rd
new file mode 100644
index 000000000..f2c1b3ced
--- /dev/null
+++ b/man/mlr_filters_ensemble.Rd
@@ -0,0 +1,12 @@
+% Generated by roxygen2: do not edit by hand
+% Please edit documentation in R/FilterEnsemble.R
+\name{mlr_filters_ensemble}
+\alias{mlr_filters_ensemble}
+\alias{FilterEnsemble}
+\title{Filter Ensemble}
+\format{
+\code{\link[R6:R6Class]{R6Class}} object inheriting from \code{\link[mlr3filters:Filter]{Filter}}.
+}
+\description{
+Implements the filte rensemble proposed in Binder (2020).
+}