Preprocessing transformers and scikit-learn compatibility

SpectroChemPy preprocessing operations can be used as transformer objects with an explicit fit() / transform() lifecycle. Feature-wise scalers learn statistics from a training set and reuse them on new data. Sample-local operations such as SNV and spectral normalization compute each observation’s statistics during transform() and do not support a safe inverse without result-attached per-observation state.

The transformers also expose get_params() / set_params() so they work with sklearn.base.clone() when scikit-learn is installed.

Load data

import numpy as np

import spectrochempy as scp

dataset = scp.read_omnic("irdata/nh4y-activation.spg")
region = dataset[:, 4000.0:2000.0]

Train / test split

We keep the first 40 spectra for training and the rest for testing.

X_train = region[:40]
X_test = region[40:]

Stateful scaling with AutoscaleTransformer

fit_transform() learns the mean and std on train (per wavenumber, dim='y'), then scales train. transform() later applies the same parameters to test — no data leakage.

scaler = scp.AutoscaleTransformer(dim="y")
X_train_scaled = scaler.fit_transform(X_train)
X_test_scaled = scaler.transform(X_test)

_ = X_train_scaled.plot(title="Train set (autoscaled)")
_ = X_test_scaled.plot(title="Test set (same scaler)")

• 
• 

Inspecting learned parameters

The transformer stores the statistics as NumPy arrays.

print("Mean shape:", scaler.mean_.shape)
print("Std shape:  ", scaler.std_.shape)

Mean shape: (1, 2075)
Std shape:   (1, 2075)

Parameter inspection with get_params / set_params

Transformers follow scikit-learn parameter conventions.

print("Original params:", scaler.get_params())

# Change the target dimension
scaler.set_params(dim="x")
print("After set_params:", scaler.get_params())

Original params: {'dim': 'y'}
After set_params: {'dim': 'x'}

sklearn.base.clone compatibility

When scikit-learn is available, transformers can be cloned exactly like any sklearn estimator.

try:
    from sklearn.base import clone

    original = scp.AutoscaleTransformer(dim="x")
    original.fit(X_train)

    cloned = clone(original)
    print("Clone has same params:", cloned.get_params() == original.get_params())
    print("Clone is not fitted yet:", not cloned._fitted)
except ImportError:
    print("scikit-learn not installed — clone example skipped")

Clone has same params: True
Clone is not fitted yet: True

Inverse transform

inverse_transform() restores the original absorbance units. This is useful when interpreting model predictions (e.g., converting PCA scores back to original space). Since changing parameters invalidates fitted transformer state, set the intended parameters and fit before inverting.

scaler.set_params(dim="y")
scaler.fit(X_train)
X_train_restored = scaler.inverse_transform(X_train_scaled)
print(
    "Restored data matches original?",
    np.allclose(X_train_restored.data, X_train.data),
)

Restored data matches original? True

Uncomment the following line to display all figures when running the script directly with Python.

# scp.show()