بتل کد Battle Codeدارد باز می‌شود…

Step 2: Accuracy per Digit چالش 2 از جلسه‌ی 64، متوسط

قدم ۲: تابع accuracy آماده است. داخل حلقه‌ی per_class را کامل کن: برای رقم d بشمار چند بار در true آمده (seen) و چند بارش در pred هم همان d است (right). خروجی، بعد از سه خط قدم ۱، باید این باشد:
Accuracy: 22/30 = 73.3%
Per digit: 0:100% 1:67% 2:100% 3:67% 4:100% 5:33% 6:100% 7:100% 8:0% 9:67%
Weakest digit: 8 (0%)

card = {
    "name": "Digit Reader",
    "kind": "neural network 25-16-10 (sessions 41-47)",
    "data": "5x5 digits: 200 noisy copies to train, 30 new ones drawn by classmates to test",
    "train": 200,
    "test": 30,
}
total = card["train"] + card["test"]
print("Model:", card["name"], "-", card["kind"])
print("Data:", card["data"])
print(f"Split: {card['train']} train / {card['test']} test ({round(100 * card['test'] / total)}% test)")


true = [0, 1, 2, 3, 4, 5, 6, 7, 8, 9] * 3
pred = [0, 1, 2, 3, 4, 5, 6, 7, 3, 9,
        0, 7, 2, 3, 4, 6, 6, 7, 3, 8,
        0, 1, 2, 8, 4, 6, 6, 7, 3, 9]


def accuracy(true, pred):
    correct = 0
    for t, p in zip(true, pred):
        if t == p:
            correct += 1
    return correct, len(true)


def per_class(true, pred):
    result = {}
    for d in sorted(set(true)):
        seen = 0
        right = 0
        # count how often digit d appears in true (seen)
        # and how often it was also predicted as d (right)
        result[d] = round(100 * right / seen)
    return result


correct, n = accuracy(true, pred)
print(f"Accuracy: {correct}/{n} = {round(100 * correct / n, 1)}%")
classes = per_class(true, pred)
print("Per digit:", " ".join(f"{d}:{p}%" for d, p in classes.items()))
weakest = min(classes, key=classes.get)
print(f"Weakest digit: {weakest} ({classes[weakest]}%)")

راهنما و جواب این چالش · برگشت به درس جلسه‌ی 64: ارائه و پورتفولیو · چالش بعدی: Step 3: The Most Common Mistake

بیشتر بخوان: تابع در پایتون: آموزش def، پارامتر و return با مثال