• mlr3 Manual
  • Quickstart
  • 1 Introduction and Overview
  • 2 Basics
    • 2.1 Quick R6 Intro for Beginners
    • 2.2 Tasks
      • 2.2.1 Task Types
      • 2.2.2 Task Creation
      • 2.2.3 Predefined tasks
      • 2.2.4 Task API
        • 番羽土墙软件 Retrieving Data
        • 2.2.4.2 Roles (Rows and Columns)
        • 2.2.4.3 Task Mutators
      • 2.2.5 Plotting Tasks
    • 2.3 Learners
      • 2.3.1 Predefined Learners
    • 2.4 Train and Predict
      • 2.4.1 Creating Task and Learner Objects
      • 2.4.2 Setting up the train/test splits of the data
      • 2.4.3 Training the learner
      • 2.4.4 Predicting
      • 2.4.5 Changing the Predict Type
      • 2.4.6 Plotting Predictions
      • 番羽土啬吧 Performance assessment
    • 2.5 Resampling
      • 2.5.1 Settings
      • 2.5.2 Instantiation
      • 番羽土墙软件 Execution
      • 2.5.4 Custom resampling
      • 墙软件 Plotting Resample Results
    • 2.6 Benchmarking
      • 2.6.1 Design Creation
      • 2.6.2 Execution and Aggregation of Results
      • 番羽土墙 Plotting Benchmark Results
      • 2.6.4 Extracting ResampleResults
      • 2.6.5 Converting and Merging ResampleResults
    • 2.7 Binary classification
      • 2.7.1 ROC Curve and Thresholds
      • 2.7.2 Threshold Tuning
  • 3 Model Optimization
    • 3.1 Hyperparameter Tuning
      • 3.1.1 The TuningInstance Classes
      • 3.1.2 The Tuner Class
      • 3.1.3 Triggering the Tuning
      • 3.1.4 Automating the Tuning
    • 3.2 Feature Selection / Filtering
      • 3.2.1 Filters
      • 3.2.2 Calculating filter values
      • 3.2.3 Variable Importance Filters
      • 3.2.4 Ensemble Methods
      • 3.2.5 Wrapper Methods
    • 3.3 Nested Resampling
      • 番羽土啬吧 Execution
      • 3.3.2 Evaluation
  • 4 Pipelines
    • 4.1 The Building Blocks: PipeOps
    • 4.2 The Pipeline Operator: %>>%
    • 4.3 Nodes, Edges and Graphs
    • 4.4 Modeling
      • 4.4.1 Setting Hyperparameters
      • 墙软件 Tuning
    • 4.5 Non-Linear Graphs
      • 4.5.1 Branching & Copying
      • 4.5.2 Model Ensembles
        • 4.5.2.1 Bagging
        • 番羽土墙 Stacking
        • 4.5.2.3 Multilevel Stacking
    • 4.6 Special Operators
      • 4.6.1 Imputation: PipeOpImpute
      • 4.6.2 Feature Engineering: PipeOpMutate
      • 番羽土啬吧 Training on data subsets: PipeOpChunk
      • 墙软件 Feature Selection: 墙软件 and PipeOpSelect
    • 4.7 In-depth look into mlr3pipelines
      • 番羽土墙软件 What’s the Point
      • 4.7.2 PipeOp: Pipeline Operators
        • 4.7.2.1 Why the $state
        • 番羽土啬吧 Where to Get PipeOps
      • 4.7.3 PipeOp Channels
        • 4.7.3.1 Input Channels
        • 4.7.3.2 Output Channels
        • 4.7.3.3 Channel Configuration
      • 4.7.4 番羽土啬吧: Networks of PipeOps
        • 4.7.4.1 Basics
        • 4.7.4.2 Networks
        • 4.7.4.3 Syntactic Sugar
        • 4.7.4.4 PipeOp IDs and ID Name Clashes
      • 4.7.5 Learners in Graphs, Graphs in Learners
        • 4.7.5.1 PipeOpLearner
        • 4.7.5.2 GraphLearner
      • 番羽土墙软件 Hyperparameters
  • 5 Technical
    • 5.1 Parallelization
    • 5.2 Error Handling
      • 5.2.1 Encapsulation
      • 5.2.2 Fallback learners
    • 5.3 Database Backends
      • 番羽土啬吧 Use Case: NYC Flights
      • 5.3.2 Preprocessing with 墙软件
      • 5.3.3 DataBackendDplyr
      • 番羽土墙 Model fitting
      • 5.3.5 Cleanup
    • 5.4 Parameters (using paradox)
      • 5.4.1 Reference Based Objects
      • 5.4.2 Defining a Parameter Space
        • 5.4.2.1 Single Parameters
        • 5.4.2.2 Parameter Sets
        • 5.4.2.3 Vector Parameters
      • 5.4.3 Parameter Sampling
        • 5.4.3.1 Parameter Designs
        • 5.4.3.2 Grid Design
        • 番羽土墙 Random Sampling
        • 5.4.3.4 Generalized Sampling: The Sampler Class
      • 5.4.4 Parameter Transformation
        • 5.4.4.1 Transformation between Types
    • 5.5 Logging
      • 5.5.1 Changing mlr3 logging levels
      • 墙软件 Redirecting output
  • 6 Extending
    • 6.1 Adding new Learners
      • 6.1.1 Meta-information
      • 6.1.2 Train function
      • 墙软件 Predict function
      • 6.1.4 Control objects/functions of learners
      • 6.1.5 Testing the learner
        • 6.1.5.1 Train and Predict
        • 6.1.5.2 Checking Parameters
      • 番羽土墙 Learner FAQ
    • 6.2 Adding new Measures
    • 6.3 Adding new PipeOps
      • 6.3.1 General Case Example: PipeOpCopy
        • 6.3.1.1 First Steps: Inheriting from PipeOp
        • 6.3.1.2 Channel Definitions
        • 番羽土墙 Train and Predict
        • 番羽土墙软件 Putting it Together
      • 6.3.2 Special Case: Preprocessing
        • 6.3.2.1 Example: PipeOpDropNA
        • 番羽土啬吧 Example: PipeOpScaleAlways
      • 6.3.3 Special Case: Preprocessing with Simple Train
        • 6.3.3.1 Example: 墙软件
        • 6.3.3.2 Example: PipeOpScaleAlwaysSimple
      • 6.3.4 Hyperparameters
        • 6.3.4.1 Hyperparameter Example: PipeOpScale
  • 7 Special Tasks
    • 7.1 Survival Analysis
      • 7.1.1 TaskSurv
      • 番羽土墙 Predict Types - crank, lp, and distr