Skip to content

FILE: docs/api/numpy.md

NumPy API Reference

Functions for array operations, normalization, safe division, and reproducibility.

normalize(arr, method='minmax', axis=0, feature_range=(0, 1), clip=False)

Normalize or standardize an array.

Parameters: - arr – numpy array or pandas Series - method – 'minmax', 'zscore', or 'robust' - axis – 0 for columns, 1 for rows (2D arrays) - feature_range – output range for minmax method - clip – clip output to feature_range (minmax only)

Returns: Normalized numpy array

Example: 

norm_arr = dk.normalize(arr)
norm_arr = dk.normalize(arr, method='zscore')
norm_arr = dk.normalize(arr, method='minmax', feature_range=(-1, 1))

safe_divide(numerator, denominator, fill=0.0, nan_safe=True)

Divide arrays safely, handling zeros and NaNs.

Parameters: - numerator – array-like or scalar - denominator – array-like or scalar - fill – value to use where denominator is 0 or NaN - nan_safe – treat NaN denominators as 0 → fill

Returns: numpy array of results

Example: 

result = dk.safe_divide(prices, quantities, fill=0.0)
result = dk.safe_divide(a, b, fill=np.nan, nan_safe=False)

set_seed(n, torch=True, tf=True)

Set all random seeds for reproducibility.

Parameters: - n – integer seed value - torch – seed PyTorch if available - tf – seed TensorFlow if available

Returns: None (prints confirmation)

Example: 

dk.set_seed(42)
dk.set_seed(123, torch=False, tf=False)

assert_shape(arr, expected=None, fix=None, name='array')

Check array shape with optional auto‑fix.

Parameters: - arr – numpy array - expected – expected shape tuple (or None) - fix – 'flatten', 'squeeze', 'expand', 'reshape_col', 'reshape_row' - name – name for error messages

Returns: Array (possibly reshaped) or raises ValueError

Example: 

arr = dk.assert_shape(arr, expected=(100, 1), fix='expand')
dk.assert_shape(arr, fix='flatten')
print(dk.assert_shape(arr))  # prints shape