Spaces:
Running
Running
File size: 2,293 Bytes
9d23e11 5b82f0a b36f3d0 5b82f0a b36f3d0 14f6420 5b82f0a 9d23e11 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 | """Normalization helpers for mapping raw uncertainty to a 0..100 display score."""
from __future__ import annotations
from dataclasses import dataclass
import json
from pathlib import Path
import numpy as np
@dataclass(frozen=True)
class QuantileNormalizer:
"""Map a non-negative raw uncertainty score onto a 0..100 display scale."""
boundaries: tuple[float, ...]
def __post_init__(self) -> None:
if len(self.boundaries) < 2:
raise ValueError("Quantile boundaries must contain at least min and max.")
if any(boundary < 0.0 for boundary in self.boundaries):
raise ValueError("Quantile boundaries must be non-negative.")
if tuple(self.boundaries) != tuple(sorted(self.boundaries)):
raise ValueError("Quantile boundaries must be sorted in ascending order.")
if self.boundaries[0] == self.boundaries[-1]:
raise ValueError("Quantile boundaries must span a non-zero range.")
def normalize(self, raw_score: float) -> float:
"""Map a raw uncertainty value onto a 0..100 scale with clamping."""
if raw_score < 0.0:
raise ValueError("raw_score must be non-negative.")
quantile_positions = np.linspace(0.0, 100.0, num=len(self.boundaries))
normalized_score = np.interp(raw_score, self.boundaries, quantile_positions)
return float(np.clip(normalized_score, 0.0, 100.0))
def band(self, raw_score: float) -> str:
"""Map a raw uncertainty value onto a coarse display band."""
normalized_score = self.normalize(raw_score)
if normalized_score < (100.0 / 3.0):
return "low"
if normalized_score < (200.0 / 3.0):
return "mid"
return "high"
def load_quantile_normalizer(path: str | Path) -> QuantileNormalizer:
"""Load a normalizer from a JSON file containing ordered boundaries."""
config_path = Path(path)
with config_path.open("r", encoding="utf-8") as config_file:
payload = json.load(config_file)
boundaries = payload.get("boundaries")
if not isinstance(boundaries, list):
raise ValueError("Normalization config must contain a 'boundaries' list.")
return QuantileNormalizer(boundaries=tuple(float(boundary) for boundary in boundaries))
|