Tutte le sfide
Facile 0 risolte 0% di accettazione

Categorical Cross-Entropy Loss

Measure the loss of a multi-class classifier from its predicted probabilities and the true classes.

Cosa imparerai

  • Score predicted probabilities against true classes
  • Average the loss over samples
  • Avoid log(0) with clipping

Problema

Mostra traduzione

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.

Vincoli

1 ≤ samples ≤ 1,000, 2 ≤ classes ≤ 100

0 ≤ each probability ≤ 1

0 ≤ labels[i] < number of classes

Esempi

Esempio 1
Input:
probabilities = [
    [0.7, 0.2, 0.1],
    [0.1, 0.8, 0.1]
]
labels = [0, 1]

Output atteso:
expected = 0.289909
Input:
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};

Output atteso:
double expected = 0.289909;
Input:
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};

Output atteso:
double expected = 0.289909;

Spiegazione

The correct classes got 0.7 and 0.8: -(ln(0.7) + ln(0.8)) / 2 = -(-0.356675 - 0.223144) / 2 ≈ 0.289909.

Esempio 2
Input:
probabilities = [
    [1.0, 0.0]
]
labels = [1]

Output atteso:
expected = 16.118096
Input:
std::vector<std::vector<double>> probabilities = {
    {1.0, 0.0}
};
std::vector<long long> labels = {1};

Output atteso:
double expected = 16.118096;
Input:
double[][] probabilities = new double[][] {
    new double[] {1.0, 0.0}
};
long[] labels = new long[] {1};

Output atteso:
double expected = 16.118096;

Spiegazione

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.

Come affrontarlo

Prova prima da solo, poi apri un suggerimento alla volta.

Suggerimento 1

Look up the probability assigned to each sample’s correct class, then average the negative logs of those values.

Suggerimento 2

Clamp very small probabilities upward before taking logs to avoid undefined results.

Suggerimento 3

Only the labeled class matters for each row, so ignore the other class probabilities when computing the loss.

Complessità attesa

O(samples) time, O(1) extra space

Riferimenti

Questa sfida è adattata per CodeUtility. Segui la traccia e i vincoli sopra.

La migliore soluzione superata di ogni programmatore. Le più veloci, leggere e meglio valutate stanno in alto.

Ancora nessuna soluzione superata. Risolvila e prendi il primo posto.

Impara dagli altri

Soluzioni della community

Nessuna soluzione pubblicata. Risolvi questa sfida per condividere la prima.