Проект готов

.gitignore

@@ -1 +1,3 @@
-.venv/
+.venv/
+data
+.ipynb_checkpoints/

plot_sgd_comparison copy.py

@@ -0,0 +1,108 @@
+import torch
+from torchvision import datasets, transforms
+import numpy as np
+import matplotlib.pyplot as plt
+
+from sklearn.linear_model import (
+    LogisticRegression,
+    PassiveAggressiveClassifier,
+    Perceptron,
+    SGDClassifier,
+)
+from sklearn.model_selection import train_test_split
+
+# -----------------------------
+# 1. Загрузка и подготовка данных
+# -----------------------------
+
+transform = transforms.Compose([
+    transforms.ToTensor(),         # Преобразуем в тензор
+    lambda x: x.view(-1).numpy()   # Преобразуем изображение в одномерный массив
+])
+
+# Загружаем EMNIST (Letters): содержит буквы A-Z
+train_dataset = datasets.EMNIST(
+    root='./data', split='letters', train=True, download=True, transform=transform
+)
+test_dataset = datasets.EMNIST(
+    root='./data', split='letters', train=False, download=True, transform=transform
+)
+
+# Объединяем train и test
+X_train = [x for x, y in train_dataset]
+y_train = [y - 1 for x, y in train_dataset]  # метки от 1 до 26 -> делаем 0..25
+
+X_test = [x for x, y in test_dataset]
+y_test = [y - 1 for x, y in test_dataset]
+
+# Объединяем всё в один набор
+X = np.array(X_train + X_test)
+y = np.array(y_train + y_test)
+
+# 🔍 Ограничиваем данные до ~1800 образцов (как в digits())
+SAMPLE_LIMIT = 1800
+X = X[:SAMPLE_LIMIT]
+y = y[:SAMPLE_LIMIT]
+
+print("Данные загружены:", X.shape, y.shape)
+
+# -----------------------------
+# 2. Настройка моделей
+# -----------------------------
+
+heldout = [0.95, 0.90, 0.75, 0.50, 0.01]  # доли тестовой выборки
+rounds = 10  # число повторений для усреднения
+
+classifiers = [
+    ("SGD", SGDClassifier(max_iter=110)),
+    ("ASGD", SGDClassifier(max_iter=110, average=True)),
+    ("Perceptron", Perceptron(max_iter=110)),
+    (
+        "Passive-Aggressive I",
+        PassiveAggressiveClassifier(max_iter=110, loss="hinge", C=1.0, tol=1e-4),
+    ),
+    (
+        "Passive-Aggressive II",
+        PassiveAggressiveClassifier(
+            max_iter=110, loss="squared_hinge", C=1.0, tol=1e-4
+        ),
+    ),
+    (
+        "SAG",
+        LogisticRegression(max_iter=110, solver="sag", tol=1e-1, C=1.0e4 / X.shape[0]),
+    ),
+]
+
+xx = 1.0 - np.array(heldout)  # пропорция обучающей выборки
+
+# -----------------------------
+# 3. Обучение и оценка моделей
+# -----------------------------
+
+for name, clf in classifiers:
+    print(f"Обучение: {name}")
+    rng = np.random.RandomState(42)
+    yy = []
+    for test_size in heldout:
+        errors = []
+        for r in range(rounds):
+            X_train_part, X_test_part, y_train_part, y_test_part = train_test_split(
+                X, y, test_size=test_size, random_state=rng
+            )
+            clf.fit(X_train_part, y_train_part)
+            y_pred = clf.predict(X_test_part)
+            error_rate = 1 - np.mean(y_pred == y_test_part)
+            errors.append(error_rate)
+        yy.append(np.mean(errors))
+    plt.plot(xx, yy, label=name)
+
+# -----------------------------
+# 4. Визуализация результатов
+# -----------------------------
+
+plt.legend(loc="upper right")
+plt.xlabel("Пропорция обучающей выборки")
+plt.ylabel("Ошибка на тесте")
+plt.title("Сравнение онлайн-алгоритмов на уменьшенном EMNIST Letters")
+plt.grid(True)
+plt.show()

plot_sgd_comparison.py

@@ -0,0 +1,70 @@
+"""
+==================================
+Comparing various online solvers
+==================================
+An example showing how different online solvers perform
+on the hand-written digits dataset.
+"""
+
+# Authors: The scikit-learn developers
+# SPDX-License-Identifier: BSD-3-Clause
+
+import matplotlib.pyplot as plt
+import numpy as np
+
+from sklearn import datasets
+from sklearn.linear_model import (
+    LogisticRegression,
+    PassiveAggressiveClassifier,
+    Perceptron,
+    SGDClassifier,
+)
+from sklearn.model_selection import train_test_split
+
+heldout = [0.95, 0.90, 0.75, 0.50, 0.01]
+# Number of rounds to fit and evaluate an estimator.
+rounds = 10
+X, y = datasets.load_digits(return_X_y=True)
+
+classifiers = [
+    ("SGD", SGDClassifier(max_iter=110)),
+    ("ASGD", SGDClassifier(max_iter=110, average=True)),
+    ("Perceptron", Perceptron(max_iter=110)),
+    (
+        "Passive-Aggressive I",
+        PassiveAggressiveClassifier(max_iter=110, loss="hinge", C=1.0, tol=1e-4),
+    ),
+    (
+        "Passive-Aggressive II",
+        PassiveAggressiveClassifier(
+            max_iter=110, loss="squared_hinge", C=1.0, tol=1e-4
+        ),
+    ),
+    (
+        "SAG",
+        LogisticRegression(max_iter=110, solver="sag", tol=1e-1, C=1.0e4 / X.shape[0]),
+    ),
+]
+
+xx = 1.0 - np.array(heldout)
+
+for name, clf in classifiers:
+    print("training %s" % name)
+    rng = np.random.RandomState(42)
+    yy = []
+    for i in heldout:
+        yy_ = []
+        for r in range(rounds):
+            X_train, X_test, y_train, y_test = train_test_split(
+                X, y, test_size=i, random_state=rng
+            )
+            clf.fit(X_train, y_train)
+            y_pred = clf.predict(X_test)
+            yy_.append(1 - np.mean(y_pred == y_test))
+        yy.append(np.mean(yy_))
+    plt.plot(xx, yy, label=name)
+
+plt.legend(loc="upper right")
+plt.xlabel("Proportion train")
+plt.ylabel("Test Error Rate")
+plt.show()