Unverified50% confidenceFactExact time
A neural network without an activation function can only represent linear functions no matter how deep it is.
1
Sources
50%
Confidence
Long-term
Relevance
10/13/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThe universal approximation theorem is a purely existential conclusion that guarantees the existence of an approximation scheme but provides no tight upper bound on the number of neurons required.67% similarUnverifiedWhether real deep nonlinear networks obey the same dual phase transition pattern remains an open question, since the paper's conclusions are built on the idealized assumptions of a single hidden layer, linear activation, and isotropic Gaussian prior.66% similarUnverifiedLinear reward functions cannot express certain goals with non-Markovian properties.65% similarUnverifiedThe universal approximation theorem only guarantees the existence of such a network; it does not tell how to efficiently find the parameters, nor does it guarantee generalization ability.64% similarUnverifiedThe human brain is naturally adept at processing linear relationships, yet extremely poor at intuitively understanding exponential growth.61% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1013751API
curl https://kongchang.com/api/v1/knowledge/claims/1013751MCP
get_claim(id=1013751)