Simulations | mlr3 benchmark | Downsample analysis | iid easy task | different task | Conclusion | Stop future background workers

Introduction: role definitions | Example using AZtrees data | Cross-validation with no column roles | Cross-validation using strata | Cross-validation on polygons | Cross-validation on polygons with strata | Cross-validation with subsets, strata, and groups | Example using respiratory data | Are models consistent across sex? | Are models consistent across centers? | Conclusion

Introduction: tasks, learners, resampling | Resampling | Tasks | Learners | Define the grid of combinations | Previous method, mlr3::benchmark_grid() | Proposed method, mlr3resampling::proj_grid() | Testing | Testing one job | Test one job for each algo and data set | Running all jobs locally (small benchmarks) | Running all jobs on a cluster (large benchmarks) | Previous method, batchtools | Proposed method, mlr3resampling::proj_submit() | Results comparison | Accuracy measures | Computation time | New features | New edit_learner() method for quick testing | New save_learner() method for model interpretation | New Task down-sampling | Conclusion | Stop future background workers

Introduction | Example | Demonstration with test project | Demonstration with full benchmark | Do it yourself | Comparison with batchtools | Conclusion

Simulations | Comparing training on Same/Other/All subsets | Downsample to see how many train data are required for good accuracy overall | Reproducibility | Reproducing K-fold CV for largest train size | Reproducing each split | Downsample to sizes of other sets | Use with auto_tuner on a task with stratification and grouping | Conclusions | Arizona trees data | What is a group? | What is a subset? | Cross-validation | Benchmark and test error computation | Conclusion | Session info

Asymptotic time/memory measurement of different R expressions | predict method | Custom asymptotic references | Exercises for the reader

Polygon methods | SO post about stats | serialize issue | changepoint cost minima | LOPART ROC curve | white or black text on colored background | odd qp labels for timings figure | two dlgrobs | dl.add

Simple regex example | Dynamic programming algorithms for change-point detection

Allocation and length | Comparison with bench::press | Complexity class estimation with atime | Matrix Multiplication, 90% sparsity | Matrix multiplication, linear number of non-zeros | Matrix multiplication, linear and quadratic number of non-zeros | Conclusion

Interpolate at seconds.limit using predict method | atime_grid to compare different engines

nc::field for reducing repetition | nc::quantifier for fewer parentheses | nc::alternatives for simplified alternation | nc::alternatives_with_shared_groups for alternatives with identical named sub-pattern groups

Capture first match in several subjects | Capture all matches in a single subject | Reshape a data table with regularly named columns | Reading regularly named data files | Helper functions for defining complex pattterns

Example 0: iris data, one file per species | Example 1: four files, two capture groups, custom read function | Why not base R? | Example 3: Hive partition file names | Example 4: pattern with two more capture groups | Example 5: parsing non-CSV data | Conclusion

Match all tracks in the text file | Match all fields in each track | Match all tracks and some fields with one regex | Parsing SweeD output files

Capture first match in each character vector element | View generated regex | Error/NA if any subjects do not match | Other regex engines | Capture first match from one or more character columns in a table | Complex example: fsa file names

Reshaping several input columns into a single output column | Reshaping several input columns into multiple output columns | Melting WHO data with a more complex pattern | Melting monthly Eurostat data | Melting into multiple output columns with missing input columns | Multiple output columns of different types

Are sepals larger or smaller than petals in the iris data? | Comparing dimensions in iris data | Do columns need to be sorted? | Melting into a single output column, who data | Melting a wider iris back to original | Lots of column types | family data | histogram of iris | single output column simple type conversion | Multiple output columns of different types

Related work | Simulate a noisy integer vector with changes | Segment a vector of integers | Segment a data frame | Write the file yourself | Computing the model with a given number of peaks

Introduction | Datasets | Anatomy of a Data Visualization | Making Animints | Using showSelected | Using clickSelects | Linked Plots | Conclusion | Footnotes

No label constraints

Simple data set where depmix errors for nstates=2 | Data set where nstates=5 errors | Fit several data sequences | How to constrain common variance parameter? | Forward-backward algorithm speed comparison | Viterbi algorithm speed comparison

Neuroblastoma data | R implementation | Synthetic data

Comparing cross-validation methods | How many times is sufficient?

Comparisons | Comparing loss/parameters to changepoint | Penalized model selection with binsegRcpp and changepoint

Simple data set where depmix errors for nstates=2

Introduction | Details | Conclusion

Introduction | Weighted ROC curve | Weighted AUC