Skip to content

Preprocess

The Preprocess class is a core building block in AutonFeat. This enables users to define custom preprocessors that can be applied to the signal before extracting features.

Bases: object

Represents a preprocessor to apply to a signal.

__call__(signal, *args, **kwargs)

Apply the preprocessor to the signal provided.

Parameters:

Name Type Description Default
signal ndarray

The signal window to apply the preprocessor to.

 required
*args Any

Additional arguments to pass to the preprocessor.

 ()
**kwargs Any

Additional keyword arguments to pass to the preprocessor.

 {}

Returns:

Type Description
ndarray

The preprocessed signal.

Raises:

Type Description
NotImplementedError

If the preprocessor is not implemented.

__init__(name='Not specified')

Initialize a new preprocessor.

__repr__()

Get the string representation of the preprocessor.

Returns:

Type Description
str

The string representation of the preprocessor.

__str__()

Get the string representation of the preprocessor.

Returns:

Type Description
str

The string representation of the preprocessor.

get_name()

Get the name of the preprocessor.

Returns:

Type Description
str

The name of the preprocessor.

set_name(name)

Set the name of the preprocessor.

Parameters:

Name Type Description Default
name str

The new name of the preprocessor.

 required

Examples

In this example, we define a custom preprocessor that shifts the signal by some delta value.

Define Custom Preprocessor

We define a custom preprocessor that performs this computation in the following way.

import numpy as np
from typing import Union, Callable
from autonfeat.core import Preprocess

class DeltaPreprocessor(Preprocess):
    """
    Preprocess the signal by shifting the `signal` by some `delta` value.
    """
    def __init__(self, name: str = "Delta") -> None:
        super().__init__(name=name)

    def __call__(self, signal: np.ndarray, delta: Union[int, float, np.int_, np.float_], where: Callable[[Union[int, float, np.int_, np.float_]], Union[bool, np.bool_]] = lambda x: not np.isnan(x)) -> np.ndarray:
        where_fn = np.vectorize(where)
        # Compute mask and multiply by distribution shift along axis
        mask = where_fn(x)
        shift = mask * delta
        return x - shift

Apply Custom Preprocessor

The preprocessor can then be applied to a signal \(x\) as follows:

# Define the signal
x = np.array([1, 2, 3, 4, 5, 6, 7, 8, 9])

# Define the delta
delta = 1
delta_preprocessor = DeltaPreprocessor()

# Apply the preprocessor
processed_signal = delta_preprocessor(x, delta=delta)

# See the result
print(processed_signal)

[0 1 2 3 4 5 6 7 8]

If you enjoy using AutonFeat, please consider starring the repository ⭐️.