Problema
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.
Restricciones
1 ≤ samples ≤ 1,000, 2 ≤ classes ≤ 100
0 ≤ each probability ≤ 1
0 ≤ labels[i] < number of classes
Ejemplos
Entrada:
probabilities = [
[0.7, 0.2, 0.1],
[0.1, 0.8, 0.1]
]
labels = [0, 1]
Salida esperada:
expected = 0.289909
Entrada:
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};
Salida esperada:
double expected = 0.289909;
Entrada:
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};
Salida esperada:
double expected = 0.289909;
Explicación
The correct classes got 0.7 and 0.8: -(ln(0.7) + ln(0.8)) / 2 = -(-0.356675 - 0.223144) / 2 ≈ 0.289909.
Entrada:
probabilities = [
[1.0, 0.0]
]
labels = [1]
Salida esperada:
expected = 16.118096
Entrada:
std::vector<std::vector<double>> probabilities = {
{1.0, 0.0}
};
std::vector<long long> labels = {1};
Salida esperada:
double expected = 16.118096;
Entrada:
double[][] probabilities = new double[][] {
new double[] {1.0, 0.0}
};
long[] labels = new long[] {1};
Salida esperada:
double expected = 16.118096;
Explicación
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.
Cómo abordarlo
Inténtalo primero por tu cuenta y luego abre las pistas una a una.
Pista 1
Look up the probability assigned to each sample’s correct class, then average the negative logs of those values.
Pista 2
Clamp very small probabilities upward before taking logs to avoid undefined results.
Pista 3
Only the labeled class matters for each row, so ignore the other class probabilities when computing the loss.
Complejidad esperada
O(samples) time, O(1) extra space
Referencias
Este reto está adaptado para CodeUtility. Sigue el enunciado y las restricciones indicadas arriba.