Question at position 16 Which of following activation function can’t be used at output layer of a neural network to classify an image?TanhSigmoidIf(x>5,1,0)ReLUNone of the above单项选择题
A
Tanh
B
Sigmoid
C
If(x>5,1,0)
D
ReLU
E
None of the above

Question

Question at position 16 Which of following activation function can’t be used at output layer of a neural network to classify an image?TanhSigmoidIf(x>5,1,0)ReLUNone of the above

BlackTom AI · Accepted Answer

ReLU outputs non-negative unbounded values [0, ∞), so it cannot serve as a proper output activation for producing class probabilities or interpretable class labels directly.

1) Relevant knowledge points:
- Output activations should produce values appropriate for classification: probabilities (0–1) for binary/multi-label, or normalized scores (softmax) for multi-class.  
- Activation properties: range and differentiability matter for training and interpretation.
   - Sigmoid: range (0,1) — suitable for binary probability outputs.
   - Tanh: range (−1,1) — can be mapped to class decisions (e.g., threshold) or used with postprocessing.
   - Step function If(x>5,1,0): produces discrete {0,1} labels — non-differentiable but can be used as a final decision rule after training.
   - ReLU: range [0, ∞) — unbounded positive outputs.

2) Step-by-step reasoning:
- Sigmoid: outputs p in (0,1). For binary classification, threshold (e.g., p>0.5) gives class label. Differentiable, commonly used.
- Tanh: outputs in (−1,1). For binary tasks, mapping (e.g., sign or threshold) yields class; although less common than sigmoid, it is usable.
- If(x>5,1,0): gives discrete labels; while non-differentiable (so not suitable as the activation used during gradient-based training), it can be applied to logits as a hard decision at inference time.
- ReLU: outputs y = max(0,x). Range is [0, ∞). For classification you need probabilities or normalized scores; ReLU can produce arbitrarily large positive values and many zeros, which neither directly give probabilities nor a normalized distribution for multi-class decisions. Using ReLU alone at the final layer prevents direct interpretation as class probabilities and breaks common loss choices (cross-entropy expects probabilities/softmax).

3) Intermediate “calculations” (ranges):
- Sigmoid(x) ∈ (0,1)
- Tanh(x) ∈ (−1,1)
- If(x>5,1,0) ∈ {0,1}
- ReLU(x) ∈ [0, ∞)

4) Why ReLU is the correct (cannot be used) choice:
Because it does not produce bounded probability-like outputs and cannot yield a normalized distribution for multi-class classification; it also makes typical classification losses and interpretation invalid if used alone as the final activation.

5) Why other options are incorrect:
- Sigmoid: correct for binary classification.
- Tanh: usable with thresholding or mapping.
- If(x>5,1,0): though non-differentiable, it can serve as a hard decision rule at output/inference.
- "None of the above": incorrect since ReLU is unsuitable.

Conclusion: The correct answer is ReLU, which matches the provided answer.

Question at position 16 Which of following activation function can’t be used at output layer of a neural network to classify an image?TanhSigmoidIf(x>5,1,0)ReLUNone of the above单项选择题

类似问题

Which statements are correct regarding "Activation" Function?

Question at position 17 Which activation function is commonly used in the target layer when modeling a binary target?exponentiallogisticidentitytanh

Which of the followings is not a widely used activation function?

Question at position 13 What activation function should we use for the output layer for the MNIST dataset?SoftmaxReLU

Question at position 13 What activation function should we use for the output layer for the MNIST dataset?SoftmaxReLU

Question at position 13 What activation function should we use for the output layer for the MNIST dataset?SoftmaxReLU

Question at position 11 Which of the following computation adds nonlinearity to an ANN?Performing activation functionTaking weighed sum of inputs

Question at position 11 Which of the following computation adds nonlinearity to an ANN?Performing activation functionTaking weighed sum of inputs

更多留学生实用工具

智能学习助手

风格化写作助手

论文查重助手

文献引用助手

课堂转译助手

课堂笔记助手

Quiz搜索助手

学校历年真题

智能刷题助手

智能匹配练习

希望你的学习变得更简单