第 14 章 演習解答¶

import numpy as np

x = np.array([-2.0, -1.0, 0.0, 1.0, 2.0])

# ReLU
relu = np.maximum(0, x)
print(f"ReLU: {relu}")

# シグモイド
sigmoid = 1 / (1 + np.exp(-x))
print(f"Sigmoid: {sigmoid}")

# ソフトマックス
softmax = np.exp(x) / np.sum(np.exp(x))
print(f"Softmax: {softmax}")
print(f"Softmax合計: {np.sum(softmax)}")

import numpy as np
from tensorflow.keras.datasets import fashion_mnist
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import (
    Input, Conv2D, MaxPooling2D, Flatten, Dense
)

# データの読み込み（問題文のコード）
(X_train, y_train), (X_test, y_test) = fashion_mnist.load_data()
X_train = X_train / 255.0
X_test = X_test / 255.0
X_train = X_train.reshape(-1, 28, 28, 1)
X_test = X_test.reshape(-1, 28, 28, 1)

label_names = ["Tシャツ", "ズボン", "セーター", "ドレス", "コート",
               "サンダル", "シャツ", "スニーカー", "バッグ", "ブーツ"]

# 本節のCNNモデルと同じ構造
cnn_model = Sequential([
    Input(shape=(28, 28, 1)),
    Conv2D(32, (3, 3), activation="relu"),
    MaxPooling2D((2, 2)),
    Conv2D(64, (3, 3), activation="relu"),
    MaxPooling2D((2, 2)),
    Flatten(),
    Dense(64, activation="relu"),
    Dense(10, activation="softmax")
])
cnn_model.compile(
    optimizer="adam",
    loss="sparse_categorical_crossentropy",
    metrics=["accuracy"]
)
cnn_model.fit(X_train, y_train, epochs=5, batch_size=64,
              validation_split=0.2)

# テストデータでの評価
test_loss, test_acc = cnn_model.evaluate(X_test, y_test, verbose=0)
print(f"テスト精度: {test_acc:.4f}")

# 先頭5枚の予測結果をラベル名で表示
pred = cnn_model.predict(X_test[:5], verbose=0)
for i, p in enumerate(pred):
    print(f"{i+1}枚目: 予測 {label_names[np.argmax(p)]}, "
          f"正解 {label_names[y_test[i]]}")

from transformers import pipeline

# 1. 日本語テキストの感情分析（本章と同じ多言語モデル）
classifier_ja = pipeline(
    "sentiment-analysis",
    model="lxyuan/distilbert-base-multilingual-cased-sentiments-student"
)
texts = [
    "この本はとても勉強になりました",
    "電車が遅延して最悪でした",
    "今日は特に何もありませんでした",
]
for text in texts:
    scores = classifier_ja(text, top_k=None)
    dist = ", ".join(f"{s['label']} {s['score']:.3f}" for s in scores)
    print(f"{text} → {dist}")

# 2. 英語のテキスト生成
# モデルを指定しない場合は大きなモデルがダウンロードされるため、
# ここでは軽量な gpt2 を指定しています
generator = pipeline("text-generation", model="gpt2")
result = generator("Machine learning is", max_new_tokens=30,
                   num_return_sequences=1)
print(result[0]["generated_text"])

from tensorflow.keras.datasets import mnist
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Input, Dense

# 本章と同じデータ準備
(X_train, y_train), (X_test, y_test) = mnist.load_data()
X_train = (X_train / 255.0).reshape(-1, 784)
X_test = (X_test / 255.0).reshape(-1, 784)

# 隠れ層の構成を変えた4つのモデルを比較
architectures = {
    "本章の構成 (128, 64)": [128, 64],
    "1層 (128)": [128],
    "3層 (256, 128, 64)": [256, 128, 64],
    "小規模 (32, 16)": [32, 16],
}
for name, hidden in architectures.items():
    layers = [Input(shape=(784,))]
    layers += [Dense(n, activation="relu") for n in hidden]
    layers += [Dense(10, activation="softmax")]
    model = Sequential(layers)
    model.compile(optimizer="adam",
                  loss="sparse_categorical_crossentropy",
                  metrics=["accuracy"])
    model.fit(X_train, y_train, epochs=10, batch_size=32,
              validation_split=0.2, verbose=0)
    _, acc = model.evaluate(X_test, y_test, verbose=0)
    print(f"{name}: パラメータ数 {model.count_params():,}, "
          f"テスト精度 {acc:.4f}")