Dependency not available for matrix_profile, this feature will be disabled!
Dependency not available for matrix_profile, this feature will be disabled!
Dependency not available for matrix_profile, this feature will be disabled!
Dependency not available for matrix_profile, this feature will be disabled!
Dependency not available for matrix_profile, this feature will be disabled!
Dependency not available for matrix_profile, this feature will be disabled!
Dependency not available for matrix_profile, this feature will be disabled!
Dependency not available for matrix_profile, this feature will be disabled!
Dependency not available for matrix_profile, this feature will be disabled!
Dependency not available for matrix_profile, this feature will be disabled!
Dependency not available for matrix_profile, this feature will be disabled!
Dependency not available for matrix_profile, this feature will be disabled!
Dependency not available for matrix_profile, this feature will be disabled!
Dependency not available for matrix_profile, this feature will be disabled!
Dependency not available for matrix_profile, this feature will be disabled!
Dependency not available for matrix_profile, this feature will be disabled!
Dependency not available for matrix_profile, this feature will be disabled!
==============================================================================
tsfresh-rs benchmark
AMD64 Family 25 Model 80 Stepping 0, AuthenticAMD | Python 3.14.3 | tsfresh 0.21.2 | tsfresh-rs 0.1.0
==============================================================================

## End to end, ComprehensiveFCParameters (783 columns per series)

workload                               tsfresh   tsfresh-rs   speedup  match
10 series x 500 pts                      2.94s      0.0278s      106x  yes
20 series x 500 pts                      5.72s      0.0464s      123x  yes
50 series x 500 pts                     14.18s      0.1007s      141x  yes
100 series x 500 pts                    27.23s      0.1951s      140x  yes
20 series x 1000 pts                    15.09s      0.1284s      118x  yes
20 series x 2000 pts                    42.41s      0.4539s       93x  yes

## Threads vs processes: what n_jobs does to each

workload                n_jobs        tsfresh   tsfresh-rs   speedup
20 x 500                1               4.34s      0.2687s       16x
20 x 500                6               5.82s      0.0639s       91x
20 x 500                all cores       4.21s      0.0405s      104x
100 x 500               1              21.55s      1.2553s       17x
100 x 500               6              11.43s      0.2695s       42x
100 x 500               all cores      24.05s      0.1956s      123x

## Series length: where the reference stops being usable

points             tsfresh   tsfresh-rs   speedup
500                  0.91s      0.0197s       46x
1000                 2.70s      0.0506s       53x
2000                 8.69s      0.1836s       47x
5000                53.64s      1.0416s       51x

## EfficientFCParameters - the preset that exists to avoid the cost

  tsfresh, Efficient (777 cols)            1.77s
  tsfresh-rs, Comprehensive (783 cols)   0.0445s
  -> the full feature set here is 40x faster than the reduced one there

## Per calculator, 20 series x 500 points, single-threaded

calculator                                         cols    tsfresh  tsfresh-rs  speedup  ok
approximate_entropy                                   5    2.7106s    0.17105s      16x  y
change_quantiles                                     60    0.9165s    0.01441s      64x  y
sample_entropy                                        1    0.5419s    0.02095s      26x  y
number_cwt_peaks                                      2    0.2078s    0.00761s      27x  y
augmented_dickey_fuller                               3    0.1581s    0.04214s       4x  y
agg_linear_trend                                     48    0.1442s    0.00374s      39x  y
max_langevin_fixed_point                              1    0.1081s    0.00048s     223x  y
friedrich_coefficients                                4    0.0926s    0.00073s     127x  y
lempel_ziv_complexity                                 5    0.0610s    0.00182s      34x  y
permutation_entropy                                   5    0.0474s    0.00359s      13x  y
fourier_entropy                                       5    0.0387s    0.00083s      46x  y
ar_coefficient                                       11    0.0349s    0.00933s       4x  y
benford_correlation                                   1    0.0143s    0.00214s       7x  y
linear_trend                                          5    0.0138s    0.00025s      55x  y
cwt_coefficients                                     60    0.0134s    0.00304s       4x  y
fft_coefficient                                     400    0.0119s    0.00512s       2x  y
number_peaks                                          5    0.0102s    0.00048s      21x  y
large_standard_deviation                             19    0.0086s    0.00126s       7x  y
quantile                                              8    0.0080s    0.00180s       4x  y
autocorrelation                                      10    0.0068s    0.00077s       9x  y
TOTAL (all 74 calculators)                                 5.2057s    0.30332s      17x

agreeing within 1e-09 relative: 74/74 calculators

done
