Nonlinear Benchmark Workshop Datasets

Fill in a module description here 
tmp_dir = idb.get_default_data_root()

Wiener Hammerstein Dataset

def plot_workshop_data(dataset_function,max_sequences=3):
    train_val, test = dataset_function(always_return_tuples_of_datasets=True)
    fig, axs = plt.subplots(2, 1, figsize=(12, 8), sharex=True)
    # Plot training/validation data
    for i, data in enumerate(train_val[:max_sequences]):
        axs[0].plot(data.u, alpha=0.7, label=f'Train {i+1}')
        axs[1].plot(data.y, alpha=0.7, label=f'Train {i+1}')

    # Plot test data
    for i, data in enumerate(test[:max_sequences]):
        axs[0].plot(data.u, ls='--', alpha=0.8, label=f'Test {i+1}')
        axs[1].plot(data.y, ls='--', alpha=0.8, label=f'Test {i+1}')

    axs[0].set_title('Input Sequences')
    axs[0].set_ylabel('Amplitude')
    axs[0].legend()

    axs[1].set_title('Output Sequences')
    axs[1].set_xlabel('Sample')
    axs[1].set_ylabel('Amplitude')
    axs[1].legend()

    plt.tight_layout()
plot_workshop_data(nonlinear_benchmarks.WienerHammerBenchMark)

source

dl_wiener_hammerstein

 dl_wiener_hammerstein (save_path:pathlib.Path, force_download:bool=False,
                        save_train_valid:bool=True, split_idx:int=80000)
Type Default Details
save_path Path directory the files are written to, created if it does not exist
force_download bool False force download the dataset
save_train_valid bool True save unsplitted train and valid datasets in ‘train_valid’ subdirectory
split_idx int 80000 split index for train and valid datasets
Returns None 
dl_wiener_hammerstein(tmp_dir / 'wh' )
dl_wiener_hammerstein(tmp_dir / 'wh' ,save_train_valid=False)
result = idb.run_benchmark(
    spec=BenchmarkWH_Simulation, 
    build_model=idb._dummy_build_model
)
result['metric_score']
Building model with spec: BenchmarkWH_Simulation, seed: 274805046
247.20421449579186
result = idb.run_benchmark(
    spec=BenchmarkWH_Prediction, 
    build_model=idb._dummy_build_model
)
result['metric_score']
Building model with spec: BenchmarkWH_Prediction, seed: 1203809552
241.88528087531387

Silverbox Dataset

plot_workshop_data(nonlinear_benchmarks.Silverbox)

source

dl_silverbox

 dl_silverbox (save_path:pathlib.Path, force_download:bool=False,
               save_train_valid:bool=True, split_idx:int=50000)
Type Default Details
save_path Path directory the files are written to, created if it does not exist
force_download bool False force download the dataset
save_train_valid bool True save unsplitted train and valid datasets in ‘train_valid’ subdirectory
split_idx int 50000 split index for train and valid datasets
Returns None 
dl_silverbox(tmp_dir / 'silverbox')
result = idb.run_benchmark(
    spec=BenchmarkSilverbox_Simulation, 
    build_model=idb._dummy_build_model
)
result['metric_score']
Building model with spec: BenchmarkSilverbox_Simulation, seed: 665849708
50.27112906124001
result = idb.run_benchmark(
    spec=BenchmarkSilverbox_Prediction, 
    build_model=idb._dummy_build_model
)
result['metric_score']
Building model with spec: BenchmarkSilverbox_Prediction, seed: 1014722045
45.73670007667041

Cascaded Tanks Dataset

plot_workshop_data(nonlinear_benchmarks.Cascaded_Tanks)

source

dl_cascaded_tanks

 dl_cascaded_tanks (save_path:pathlib.Path, force_download:bool=False,
                    save_train_valid:bool=True, split_idx:int=160)
Type Default Details
save_path Path directory the files are written to, created if it does not exist
force_download bool False force download the dataset
save_train_valid bool True save unsplitted train and valid datasets in ‘train_valid’ subdirectory
split_idx int 160 split index for train and valid datasets
Returns None 
dl_cascaded_tanks(tmp_dir  / 'cascaded_tanks' )
result = idb.run_benchmark(
    spec=BenchmarkCascadedTanks_Simulation, 
    build_model=idb._dummy_build_model
)
result['metric_score']
Building model with spec: BenchmarkCascadedTanks_Simulation, seed: 2125691036
6.190198457201898
result = idb.run_benchmark(
    spec=BenchmarkCascadedTanks_Prediction, 
    build_model=idb._dummy_build_model
)
result['metric_score']
Building model with spec: BenchmarkCascadedTanks_Prediction, seed: 3952414853
6.087247931804716

EMPS Dataset

plot_workshop_data(nonlinear_benchmarks.EMPS)

source

dl_emps

 dl_emps (save_path:pathlib.Path, force_download:bool=False,
          save_train_valid:bool=True, split_idx:int=18000)
Type Default Details
save_path Path directory the files are written to, created if it does not exist
force_download bool False force download the dataset
save_train_valid bool True save unsplitted train and valid datasets in ‘train_valid’ subdirectory
split_idx int 18000 split index for train and valid datasets
Returns None 
dl_emps(tmp_dir  / 'emps')
result = idb.run_benchmark(
    spec=BenchmarkEMPS_Simulation, 
    build_model=idb._dummy_build_model
)
result['metric_score']
Building model with spec: BenchmarkEMPS_Simulation, seed: 159203475
148.94509370511423
result = idb.run_benchmark(
    spec=BenchmarkEMPS_Prediction, 
    build_model=idb._dummy_build_model
)
result['metric_score']
Building model with spec: BenchmarkEMPS_Prediction, seed: 1758391957
127.00285207332765

Noisy Wiener Hammerstein

source

dl_noisy_wh

 dl_noisy_wh (save_path:pathlib.Path, force_download:bool=False,
              save_train_valid:bool=True)

the wiener hammerstein dataset with process noise

Type Default Details
save_path Path directory the files are written to, created if it does not exist
force_download bool False force download the dataset
save_train_valid bool True save unsplitted train and valid datasets in ‘train_valid’ subdirectory
Returns None 
dl_noisy_wh(tmp_dir / 'noisy_wh' )
results = result = idb.run_benchmark(
    spec=BenchmarkNoisyWH_Simulation, 
    build_model=idb._dummy_build_model
)
results['metric_score']
Building model with spec: BenchmarkNoisyWH_Simulation, seed: 3771735968
104.1542183129001
results = result = idb.run_benchmark(
    spec=BenchmarkNoisyWH_Prediction, 
    build_model=idb._dummy_build_model
)
results['metric_score']
Building model with spec: BenchmarkNoisyWH_Prediction, seed: 421895102
81.6161143620699

Parallel Wienerhammerstein

#ToDo

F16

#ToDo

Coupled Electric Drives

plot_workshop_data(nonlinear_benchmarks.CED)

source

dl_ced

 dl_ced (save_path:pathlib.Path, force_download:bool=False,
         save_train_valid:bool=True, split_idx:int=300)
Type Default Details
save_path Path directory the files are written to, created if it does not exist
force_download bool False force download the dataset
save_train_valid bool True save unsplitted train and valid datasets in ‘train_valid’ subdirectory
split_idx int 300 split index for train and valid datasets
Returns None 
dl_ced(tmp_dir / 'ced' )
result = idb.run_benchmark(
    spec=BenchmarkCED_Simulation, 
    build_model=idb._dummy_build_model
)
result['metric_score']
Building model with spec: BenchmarkCED_Simulation, seed: 750852011
0.5816430221160263
result = idb.run_benchmark(
    spec=BenchmarkCED_Prediction, 
    build_model=idb._dummy_build_model
)
result['metric_score']
Building model with spec: BenchmarkCED_Prediction, seed: 4123995934
0.49837518028637195