Answer-first summary for fast verification
Answer: Softmax, as it assigns probabilities to each class such that their sum equals 1, making it ideal for multi-class classification problems.
Softmax is the correct choice for multi-class classification because it normalizes the output of the network to a probability distribution over predicted output classes. This means that the output values are between 0 and 1 and sum to 1, which is essential for multi-class classification tasks. ReLU (A) is typically used in hidden layers to introduce non-linearity but is not suitable for the output layer in classification tasks. TANH (B) outputs values between -1 and 1, which is not directly interpretable as probabilities. SIGMOID (D) is used for binary classification, outputting a single probability value, and is not suitable for multi-class scenarios where multiple classes are involved. The scenario does not require a custom activation function (E), making Softmax the most appropriate choice.
Author: LeetQuiz Editorial Team
Ultimate access to all questions.
No comments yet.
You are a junior Data Scientist working on a project that requires the development of a multi-class classification model using the Keras Sequential API. The model will classify images into one of 10 possible categories. Given the project's requirements for high accuracy and the need for the model to output probabilities for each class that sum to 1, which activation function should you use for the output layer? Choose the best option. (Choose one correct option)
A
ReLU, as it helps in mitigating the vanishing gradient problem and is commonly used in hidden layers of neural networks.
B
TANH, because it outputs values between -1 and 1, making it suitable for models requiring normalized outputs.
C
Softmax, as it assigns probabilities to each class such that their sum equals 1, making it ideal for multi-class classification problems.
D
SIGMOID, since it outputs a value between 0 and 1, suitable for binary classification tasks.
E
None of the above, as a custom activation function is required for this specific use case.