此問題已經被重新評分。

以下哪种关于 CNN 中池化的说法不正确？

Which of the following is NOT true for pooling in CNNs?

Question

此問題已經被重新評分。

以下哪种关于 CNN 中池化的说法不正确？

Which of the following is NOT true for pooling in CNNs?

BlackTom AI · Accepted Answer

Pooling in CNNs reduces the spatial dimensions (height and width) of feature maps while preserving the number of channels (feature maps), so it does not increase the number of features for the next layer.

1) Relevant knowledge points and concepts
- Pooling operations (max pooling, average pooling) operate spatially over each channel independently, summarizing local neighborhoods.
- Typical effect: (H, W, C) input → (H', W', C) output (H' < H, W' < W), so channels C remain unchanged.
- Pooling has no learnable parameters, provides translation invariance, reduces computation and overfitting risk, and lowers memory requirements.
- Changing the number of channels (features) is done by convolutional layers (e.g., 1x1 conv) or explicit channel-wise operations, not by standard pooling.

2) Step-by-step reasoning with example
- Start with an input tensor shape 32 × 32 × 64 (height × width × channels).
- Apply 2×2 max-pooling with stride 2: each 2×2 block per channel becomes one value.
- Output shape becomes 16 × 16 × 64. Channels remain 64; only spatial size halved.
- No parameters are learned during pooling; it simply aggregates values within local windows.

3) Intermediate steps (calculation)
- For pooling window k and stride s without padding: H' = floor((H − k)/s) + 1, W' = floor((W − k)/s) + 1.
- Using H = 32, k = 2, s = 2 → H' = floor((32 − 2)/2) + 1 = 16; similarly W' = 16. Channel count C unchanged.

4) Why the given statement is incorrect
- The statement "It is a method for increasing the number of features for the next layer" is false because pooling does not alter channel count; therefore it does not increase the number of features passed forward. Increasing channels requires convolutional layers designed to change depth.

5) Why typical other pooling-related claims would be true (brief)
- Pooling reduces spatial resolution — true.
- Pooling introduces spatial invariance — true.
- Pooling has no learnable parameters — true.
- Pooling reduces computation and overfitting — true.

Conclusion: The false statement is exactly "It is a method for increasing the number of features for the next layer." This matches the provided answer.

此問題已經被重新評分。以下哪种关于 CNN 中池化的说法不正确？ Which of the following is NOT true for pooling in CNNs?多项选择题

类似问题

以下哪种关于 CNN 中池化的说法不正确？ Which of the following is NOT true for pooling in CNNs?

What is the main purpose of pooling in text CNNs (Convolutional Neural Networks)? (Select one option from below.)

For stock A, we have 𝛽 𝑖 = 0.70. Suppose the expected market risk premium next year is 9% and the risk-free rate is 3%. What is the expected return of this stock based on the CAPM? (Please answer in % and round to 2 decimal places. If the answer is 8.057%, then in the box, write 8.06)

The following graph shows the Security Market Line in the current economy. If a stock has a β of 1.2, what is the expected return of this stock based on the CAPM? (Please answer in % and round to 2 decimal places. If the answer is 8.057%, then in the box, write 8.06)

Suppose Apple publicly announces earnings today and the announcement is above market expectations. If markets are semistrong-form efficient, then:

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

此問題已經被重新評分。 以下哪种关于 CNN 中池化的说法不正确？ Which of the following is NOT true for pooling in CNNs?多项选择题