Условие
A multi-class classifier outputs one probability per class (for example from softmax). Categorical cross-entropy punishes it for giving the correct class a low probability:
loss = -(ln(p₁) + ln(p₂) + … + ln(pₙ)) / n
where pᵢ is the probability the model gave to the correct class of sample i. To avoid ln(0), first raise any probability below 0.0000001 to 0.0000001.
Implement categorical_cross_entropy(probabilities, labels).
Input
probabilities: One row per sample, with one probability per class.
labels: The index of the correct class for each sample (0-based).
Output
Return the average loss as a decimal number.
Ограничения
1 ≤ samples ≤ 1,000, 2 ≤ classes ≤ 100
0 ≤ each probability ≤ 1
0 ≤ labels[i] < number of classes
Примеры
Ввод:
probabilities = [
[0.7, 0.2, 0.1],
[0.1, 0.8, 0.1]
]
labels = [0, 1]
Ожидаемый вывод:
expected = 0.289909
Ввод:
std::vector<std::vector<double>> probabilities = {
{0.7, 0.2, 0.1},
{0.1, 0.8, 0.1}
};
std::vector<long long> labels = {0, 1};
Ожидаемый вывод:
double expected = 0.289909;
Ввод:
double[][] probabilities = new double[][] {
new double[] {0.7, 0.2, 0.1},
new double[] {0.1, 0.8, 0.1}
};
long[] labels = new long[] {0, 1};
Ожидаемый вывод:
double expected = 0.289909;
Пояснение
The correct classes got 0.7 and 0.8: -(ln(0.7) + ln(0.8)) / 2 = -(-0.356675 - 0.223144) / 2 ≈ 0.289909.
Ввод:
probabilities = [
[1.0, 0.0]
]
labels = [1]
Ожидаемый вывод:
expected = 16.118096
Ввод:
std::vector<std::vector<double>> probabilities = {
{1.0, 0.0}
};
std::vector<long long> labels = {1};
Ожидаемый вывод:
double expected = 16.118096;
Ввод:
double[][] probabilities = new double[][] {
new double[] {1.0, 0.0}
};
long[] labels = new long[] {1};
Ожидаемый вывод:
double expected = 16.118096;
Пояснение
The correct class got probability 0, which is raised to 0.0000001, so the loss is -ln(0.0000001) ≈ 16.118096: a confident wrong answer is punished hard.
Как подступиться
Сначала попробуйте сами, затем открывайте подсказки по одной.
Подсказка 1
Look up the probability assigned to each sample’s correct class, then average the negative logs of those values.
Подсказка 2
Clamp very small probabilities upward before taking logs to avoid undefined results.
Подсказка 3
Only the labeled class matters for each row, so ignore the other class probabilities when computing the loss.
Ожидаемая сложность
O(samples) time, O(1) extra space
Ссылки
Это задание адаптировано для CodeUtility. Следуйте условию и ограничениям выше.