P7-3.3 用实际分类器重新比较输入表示¶

import csv
from pathlib import Path

import numpy as np
from sklearn.linear_model import LogisticRegression
from sklearn.metrics import accuracy_score

data_path = Path("docs/assets/part-07/chapter-03/p7-3-surface-patches.zh.csv")
rows = list(csv.DictReader(data_path.open(encoding="utf-8")))
pixel_columns = [name for name in rows[0] if name.startswith("pixel_")]
train_rows = [row for row in rows if row["split"] == "train"]
test_rows = [row for row in rows if row["split"] == "test"]

def pixel_matrix(selected_rows):
    return np.array([[float(row[column]) for column in pixel_columns] for row in selected_rows], dtype=float)

def column_profile(matrix):
    return matrix.reshape(len(matrix), 8, 8).mean(axis=1)

def center_band_profile(matrix):
    images = matrix.reshape(len(matrix), 8, 8)
    center = images[:, :, 3:5].mean(axis=(1, 2))
    left = images[:, :, :3].mean(axis=(1, 2))
    right = images[:, :, 5:].mean(axis=(1, 2))
    return np.column_stack([center, left, right])

raw_train, raw_test = pixel_matrix(train_rows), pixel_matrix(test_rows)
y_train = np.array([int(row["label"]) for row in train_rows])
y_test = np.array([int(row["label"]) for row in test_rows])
representations = {
    "完整 64 像素": (raw_train, raw_test),
    "8 个列平均": (column_profile(raw_train), column_profile(raw_test)),
    "3 个中心带值": (center_band_profile(raw_train), center_band_profile(raw_test)),
}
summaries, sample_records = [], []
for name, (x_train, x_test) in representations.items():
    model = LogisticRegression(max_iter=1000, random_state=7).fit(x_train, y_train)
    predictions, probabilities = model.predict(x_test), model.predict_proba(x_test)
    margins = np.abs(probabilities[:, 1] - probabilities[:, 0])
    errors, low_margin = [], []
    for row, actual, predicted, probability, margin in zip(test_rows, y_test, predictions, probabilities, margins):
        sample_records.append({"表示": name, "样本": row["sample"], "实际": int(actual), "预测": int(predicted),
                               "类别概率": [round(float(value), 3) for value in probability], "置信差": round(float(margin), 3)})
        if actual != predicted: errors.append(row["sample"])
        if margin < .25: low_margin.append(row["sample"])
    summaries.append({"表示": name, "训练输入形状": tuple(x_train.shape), "评估准确率": round(float(accuracy_score(y_test, predictions)), 3),
                      "错误样本": errors, "低置信度样本": low_margin})
print("表示运行摘要 =")
for summary in summaries: print(summary)
print("评估样本记录 =")
for record in sample_records: print(record)

完整 64 像素：训练形状 (12, 64)，评估准确率 0.75，错误为弱划痕，低置信度为弱划痕。
8 个列平均：训练形状 (12, 8)，评估准确率 0.75，错误为弱划痕，四个样本都低置信度。
3 个中心带值：训练形状 (12, 3)，评估准确率 0.75，错误为弱划痕，四个样本都低置信度。