Hello 教えて.AI !!

import numpy as np from sklearn.datasets import load_breast_cancer from sklearn.model_selection import train_test_split import matplotlib.pyplot as plt

データセットの読み込み

breast_cancer = load_breast_cancer() X, y = breast_cancer.data, breast_cancer.target

データセットの分割

X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)

ハイパーパラメータの設定

weight_init_std = 0.05 max_epochs = 20 batch_size = 32 iter_per_epoch = max(len(X_train) / batch_size, 1) epoch_cnt = 0

シグモイド関数

def sigmoid(x): return 1 / (1 + np.exp(-x))

ReLU関数

def relu(x): return np.maximum(0, x)

ソフトマックス関数

def softmax(x): exp_x = np.exp(x) return exp_x / np.sum(exp_x, axis=1, keepdims=True)

交差エントロピー誤差

def cross_entropy_error(y_pred, y_true): delta = 1e-7 return -np.mean(np.sum(y_true * np.log(y_pred + delta), axis=1))

正解率、適合率、再現率、F値の算出

def calculate_metrics(y_pred, y_true): y_pred_labels = np.argmax(y_pred, axis=1) y_true_labels = np.argmax(y_true, axis=1) accuracy = np.mean(y_pred_labels == y_true_labels)

1true_positive = np.sum((y_pred_labels == 1) & (y_true_labels == 1))
2false_positive = np.sum((y_pred_labels == 1) & (y_true_labels == 0))
3false_negative = np.sum((y_pred_labels == 0) & (y_true_labels == 1))
4
5precision = true_positive / (true_positive + false_positive)
6recall = true_positive / (true_positive + false_negative)
7f1_score = 2 * (precision * recall) / (precision + recall)
8
9return accuracy, precision, recall, f1_score

ニューラルネットワーククラス

class MLP: def init(self, input_size, hidden_size, output_size): self.params = {} self.params['W1'] = weight_init_std * np.random.randn(input_size, hidden_size) self.params['b1'] = np.zeros(hidden_size) self.params['W2'] = weight_init_std * np.random.randn(hidden_size, output_size) self.params['b2'] = np.zeros(output_size)

1def forward(self, x):
2    z1 = np.dot(x, self.params['W1']) + self.params['b1']
3    a1 = relu(z1)
4    z2 = np.dot(a1, self.params['W2']) + self.params['b2']
5    y_pred = softmax(z2)
6    return y_pred
7
8def backward(self, x, y_true, y_pred):
9    batch_size = x.shape[0]
10    
11    # 出力層の誤差
12    delta2 = (y_pred - y_true) / batch_size
13    
14    # 重みとバイアスの勾配
15    grads = {}
16    grads['W2'] = np.dot(self.hidden_layer.T, delta2)
17    grads['b2'] = np.sum(delta2, axis=0)
18    
19    # 隠れ層の誤差
20    delta1 = np.dot(delta2, self.params['W2'].T) * (self.hidden_layer > 0)
21
22    # 重みとバイアスの勾配
23    grads['W1'] = np.dot(x.T, delta1)
24    grads['b1'] = np.sum(delta1, axis=0)
25    
26    return grads
27
28def train(self, x, y_true):
29    # 順伝播
30    y_pred = self.forward(x)
31    
32    # バックプロパゲーション
33    grads = self.backward(x, y_true, y_pred)
34    
35    # パラメータの更新
36    for key in self.params.keys():
37        self.params[key] -= learning_rate * grads[key]
38
39def predict(self, x):
40    y_pred = self.forward(x)
41    return np.argmax(y_pred, axis=1)

多層ニューラルネットワークモデルの作成

model = MLP(input_size=X_train.shape[1], hidden_size=50, output_size=2)

学習

train_loss_list = [] train_acc_list = [] test_acc_list = []

for epoch in range(max_epochs): # エポックごとにシャッフル idx = np.random.permutation(len(X_train)) X_train = X_train[idx] y_train = y_train[idx]

1for i in range(0, len(X_train), batch_size):
2    # バッチデータの取得
3    X_batch = X_train[i:i+batch_size]
4    y_batch = y_train[i:i+batch_size]
5
6    # バッチデータで学習
7    model.train(X_batch, y_batch)
8    
9# エポックごとの評価
10train_pred = model.forward(X_train)
11test_pred = model.forward(X_test)
12train_loss = cross_entropy_error(train_pred, y_train)
13train_acc, _, _, _ = calculate_metrics(train_pred, y_train)
14test_acc, _, _, _ = calculate_metrics(test_pred, y_test)
15
16train_loss_list.append(train_loss)
17train_acc_list.append(train_acc)
18test_acc_list.append(test_acc)
19
20epoch_cnt += 1
21if epoch_cnt >= iter_per_epoch:
22    print("Epoch:", epoch+1)
23    print("Train Loss:", train_loss)
24    print("Train Accuracy:", train_acc)
25    print("Test Accuracy:", test_acc)
26    epoch_cnt = 0

グラフの描画

plt.plot(range(1, max_epochs+1), train_loss_list, label='Train Loss') plt.xlabel('Epoch') plt.ylabel('Loss') plt.legend() plt.show()

plt.plot(range(1, max_epochs+1), train_acc_list, label='Train Accuracy') plt.plot(range(1, max_epochs+1), test_acc_list, label='Test Accuracy') plt.xlabel('Epoch') plt.ylabel('Accuracy') plt.legend() plt.show()