Unverified50% confidenceFactExact time
Previously the community generally believed that local algorithms (such as Markov chains) perform far worse than global algorithms in tensor PCA, and it was only known that under random initialization they require a polynomially higher signal-to-noise ratio to succeed.
1
Sources
50%
Confidence
Long-term
Relevance
10/30/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedTensor PCA exhibits an information-computation gap: when the signal-to-noise ratio is above the algorithmic threshold, global algorithms such as spectral methods and sum-of-squares can successfully recover in polynomial time; below that threshold, lower bounds for these algorithms and for certain restricted computational models show that they must fail.74% similarUnverifiedAlthough over-parameterized networks have enough capacity to memorize random labels, the implicit regularization of gradient descent favors learning simple, consistent patterns first.72% similarUnverifiedThe study introduces trace-norm regularization and proves that among all matrices producing the same observed output, the optimizer tends to choose low-rank matrices.71% similarUnverifiedNaively approximating the infinite-dimensional functional gradient causes the algorithm to converge to the wrong location, introducing systematic bias.70% similarUnverifiedWhen one component of a system already has randomness, stacking a probabilistic component on top of it exponentially amplifies debugging difficulty.67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1114781API
curl https://kongchang.com/api/v1/knowledge/claims/1114781MCP
get_claim(id=1114781)