ML (Preprocessing)

bin_feature()¶

Map an encrypted value to its ordinal bucket index.

This cell verifies the bin_feature function mathematically against its cleartext counterpart, using a dynamically generated input set that covers positive, negative, zero, and array edge cases while respecting the [-7, 7] FHE RAM constraints.

In [ ]:

from concrete import fhe
from concrete_fhe_toolkit.ml.preprocessing import bin_feature
def test_bin_feature(value):
    return bin_feature(value, [2, 4, 6])
compiler = fhe.Compiler(test_bin_feature, {'value': 'encrypted'})
inputset = [(5,)]
circuit = compiler.compile(inputset)
circuit.encrypt_run_decrypt(*inputset[0])
print("bin_feature tests passed! (1/1)")

from concrete import fhe from concrete_fhe_toolkit.ml.preprocessing import bin_feature def test_bin_feature(value): return bin_feature(value, [2, 4, 6]) compiler = fhe.Compiler(test_bin_feature, {'value': 'encrypted'}) inputset = [(5,)] circuit = compiler.compile(inputset) circuit.encrypt_run_decrypt(*inputset[0]) print("bin_feature tests passed! (1/1)")

FHEBinner¶

Demonstrates the FHEBinner class.

This transformer is evaluated in cleartext here. Its FHE compilation is tested automatically when chained inside an FHEPipeline.

In [6]:

from concrete_fhe_toolkit.ml.preprocessing import FHEBinner

binner = FHEBinner([[18, 30, 45, 65], [10, 50]])
res = binner._transform_logic([55, 70])

assert list(res) == [3, 2]
print('FHEBinner logic tested successfully in cleartext! (1/1)')

from concrete_fhe_toolkit.ml.preprocessing import FHEBinner binner = FHEBinner([[18, 30, 45, 65], [10, 50]]) res = binner._transform_logic([55, 70]) assert list(res) == [3, 2] print('FHEBinner logic tested successfully in cleartext! (1/1)')

FHEBinner logic tested successfully in cleartext! (1/1)

---

FHEStandardScaler¶

Demonstrates the FHEStandardScaler class.

This transformer is evaluated in cleartext here. Its FHE compilation is tested automatically when chained inside an FHEPipeline.

In [3]:

from concrete_fhe_toolkit.ml.preprocessing import FHEStandardScaler

scaler = FHEStandardScaler(means=[50, 100], stds=[10, 20], scale=10)
res = scaler._transform_logic([65, 60])

assert list(res) == [15, -20]
print('FHEStandardScaler logic tested successfully in cleartext! (1/1)')

from concrete_fhe_toolkit.ml.preprocessing import FHEStandardScaler scaler = FHEStandardScaler(means=[50, 100], stds=[10, 20], scale=10) res = scaler._transform_logic([65, 60]) assert list(res) == [15, -20] print('FHEStandardScaler logic tested successfully in cleartext! (1/1)')

FHEStandardScaler logic tested successfully in cleartext! (1/1)

---

FHEMinMaxScaler¶

Demonstrates the FHEMinMaxScaler class.

This transformer is evaluated in cleartext here. Its FHE compilation is tested automatically when chained inside an FHEPipeline.

In [4]:

from concrete_fhe_toolkit.ml.preprocessing import FHEMinMaxScaler

scaler = FHEMinMaxScaler([0, 10], [200, 20], scale=100)
res = scaler._transform_logic([50, 15])

assert list(res) == [25, 50]
print('FHEMinMaxScaler logic tested successfully in cleartext! (1/1)')

from concrete_fhe_toolkit.ml.preprocessing import FHEMinMaxScaler scaler = FHEMinMaxScaler([0, 10], [200, 20], scale=100) res = scaler._transform_logic([50, 15]) assert list(res) == [25, 50] print('FHEMinMaxScaler logic tested successfully in cleartext! (1/1)')

FHEMinMaxScaler logic tested successfully in cleartext! (1/1)

---