Skip to content

Available Containerized Benchmarks

Katharina Eggensperger edited this page Apr 20, 2022 · 8 revisions

A list of all available benchmarks

For more details, we refer to the code of each benchmark in hpobench.benchmark.<family>.

benchmarks.ml

Family Name Type Note
pybnn(x) BNNOn* raw There are 4 benchmark in total (ToyFunction, BostonHousing, ProteinStructure, YearPrediction)
lr_benchmark LRBenchmark raw Works on OpenML Dataset IDs
nn_benchmark NNBenchmark raw Works on OpenML Dataset IDs
rf_benchmark RandomForestBenchmark raw Works on OpenML Dataset IDs
svm_benchmark SVMBenchmark raw Works on OpenML Dataset IDs
tabular_benchmark(x) TabularBenchmark tabular Tabular benchmarks for the benchmarks above ('lr', 'svm', 'xgb', 'rf', 'nn')
histgb_benchmark HistGBBenchmark raw Works on OpenML Dataset IDs
xgboost_old XGBoost*Benchmark raw Works with OpenML task ids. extended contains Additional Parameter `Booster
svm_benchmark_old SupportVectorMachine raw Works with OpenML task ids.

benchmarks.rl

Family Name Type Note
cartpole CartpoleFull raw Not deterministic.
cartpole(x) CartpoleReduced raw Not deterministic. Uses a smaller searchspace than full
learna_benchmark LeaRNA Not deterministic.
learna_benchmark MetaLearna Not deterministic.

benchmarks.nas

Family Name Type Note
tabular_benchmarks(+) *Benchmark tabular There are 4 benchmarks (SlizeLocalization, ProteinStructure, NavalPropulsion, ParkinsonsTelemonitoring)
tabular_benchmarks(+,x) *BenchmarkOriginal tabular Same as above, but with a higher lowest fidelity of 3
nasbench_101(+) NASCifar10*Benchmark tabular There are 3 benchmark in total (A, B, C)
nasbench_201(+) *NasBench201Benchmark tabular There are 3 benchmarks in total (Cifar10Valid, Cifar100, ImageNet). Note: We assume that there are 3 seeds per configuration (777,888,999). For configurations with less seeds we re-use of the existing ones (See #138).
nasbench_201(+,x) *NasBench201BenchmarkOriginal tabular Same as above, but with a higher lowest fidelity of 12
nasbench_1shot1(+) NASBench1shot1SearchSpace*Benchmark tabular There are 3 benchmarks in total (1,2,3)

benchmarks.surrogates

Family Name Type Note
paramnet_benchmark ParamNet*OnStepsBenchmark paramnet surrogate
paramnet_benchmark ParamNetReduced*OnStepsBenchmark paramnet surrogate
paramnet_benchmark ParamNet*OnTimeBenchmark paramnet surrogate
paramnet_benchmark(x) ParamNetReduced*OnTimeBenchmark paramnet surrogate
svm_benchmark SurrogateSVMBenchmark surrogate_svm surrogate

benchmarks.od

Family Name Type Note
od_ae ODAutoencoder raw Works on 15 datasets (annthyroid, arrhythmia, breastw, cardio, ionosphere, mammography, musk, optdigits, pendigits, pima, satellite, satimage-2, thyroid, vowels, wbc)
od_kde ODKernelDensityEstimation raw Works on 15 datasets (see above). No fidelities available
od_ocsvm ODOneClassSupportVectorMachine raw Works on 15 datasets (see above). No fidelities available

Note:

  • (x) benchmarks that have been used in the HPOBench paper, other benchmarks are considered experimental
  • (+) Tabular benchmarks might take a few minutes to load.

Note: All containers are uploaded here