lale.lib.lale.time_series_transformer module¶
- class lale.lib.lale.time_series_transformer.CorrelationMatrix[source]¶
Bases:
object
Calculate correlation coefficients matrix across all EEG channels.
- class lale.lib.lale.time_series_transformer.Eigenvalues[source]¶
Bases:
object
Take eigenvalues of a matrix, and sort them by magnitude in order to make them useful as features (as they have no inherent order).
- class lale.lib.lale.time_series_transformer.FFT[source]¶
Bases:
object
Apply Fast Fourier Transform to the last axis.
- class lale.lib.lale.time_series_transformer.FFTWithTimeFreqCorrelation(start, end, max_hz, scale_option)[source]¶
Bases:
object
Combines FFT with time and frequency correlation, taking both correlation coefficients and eigenvalues.
- class lale.lib.lale.time_series_transformer.FreqCorrelation(start, end, scale_option, with_fft=False, with_corr=True, with_eigen=True)[source]¶
Bases:
object
Correlation in the frequency domain. First take FFT with (start, end) slice options, then calculate correlation co-efficients on the FFT output, followed by calculating eigenvalues on the correlation co-efficients matrix. The output features are (fft, upper_right_diagonal(correlation_coefficients), eigenvalues) Features can be selected/omitted using the constructor arguments.
- class lale.lib.lale.time_series_transformer.Magnitude[source]¶
Bases:
object
Job: Take magnitudes of Complex data
- class lale.lib.lale.time_series_transformer.Pipeline(pipeline)[source]¶
Bases:
object
A Pipeline is an object representing the data transformations to make on the input data, finally outputting extracted features. pipeline: List of transforms to apply one by one to the input data
- class lale.lib.lale.time_series_transformer.Resample(sample_rate)[source]¶
Bases:
object
Resample time-series data.
- class lale.lib.lale.time_series_transformer.Slice(start, stop)[source]¶
Bases:
object
Job: Take a slice of the data on the last axis. Note: Slice(x, y) works like a normal python slice, that is x to (y-1) will be taken.
- class lale.lib.lale.time_series_transformer.StandardizeFirst[source]¶
Bases:
object
Scale across the first axis.
- class lale.lib.lale.time_series_transformer.StandardizeLast[source]¶
Bases:
object
Scale across the last axis.
- class lale.lib.lale.time_series_transformer.TimeCorrelation(max_hz, scale_option, with_corr=True, with_eigen=True)[source]¶
Bases:
object
Correlation in the time domain. First downsample the data, then calculate correlation co-efficients followed by calculating eigenvalues on the correlation co-efficients matrix. The output features are (upper_right_diagonal(correlation_coefficients), eigenvalues) Features can be selected/omitted using the constructor arguments.
- class lale.lib.lale.time_series_transformer.TimeFreqEigenVectors(*args, _lale_trained=False, _lale_impl=None, **kwargs)¶
Bases:
TrainedIndividualOp
Combined schema for expected data and hyperparameters.
This documentation is auto-generated from JSON schemas.
- Parameters
window_length (float, >=0.25 for optimizer, <=2 for optimizer, uniform distribution, default 1) – TODO
window_step (float, >=0.25 for optimizer, <=1 for optimizer, uniform distribution, default 0.5) – TODO
fft_min_freq (integer, not for optimizer, default 1) – TODO
fft_max_freq (integer, >=2 for optimizer, <=30 for optimizer, uniform distribution, default 24) – TODO
sampling_frequency (integer, not for optimizer, default 250) – TODO
- transform(X, y=None)¶
Transform the data.
- Parameters
X (array of items : array of items : array of items : float) – The input data to complete.
y (array) –
items : union type
integer
or string
- Returns
result – The input data to complete.
- Return type
array of items : array