Unverified50% confidenceFactExact time
If all layers perform only linear transformations, the overall effect of the network is equivalent to a single-layer linear model regardless of how many layers it has.
1
Sources
50%
Confidence
Long-term
Relevance
8/16/2026
First Seen
Sources
用VRML可视化MLP：XOR问题的3D神经网络演示
Related Claims
UnverifiedThe composition of multiple linear transformations is still a linear transformation; therefore, a multi-layer network without nonlinear activation functions has expressive power equivalent to a single-layer linear model.76% similarUnverified贝叶斯网络通过有向无环图显式建模变量间的条件依赖关系，适合捕捉因果链条67% similarUnverified激活函数赋予深度网络拟合任意复杂非线性函数的理论能力，若无激活函数网络本质上仍是线性变换66% similarUnverified线性表征假说认为神经网络中的概念以线性方向（方向向量）的形式存储在激活空间中65% similarUnverified贝叶斯网络通过有向无环图建模变量间条件依赖关系，Copula函数将多个边缘分布耦合为联合分布，两者对高维数据的扩展性较差64% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/756912API
curl https://kongchang.com/api/v1/knowledge/claims/756912MCP
get_claim(id=756912)