22 problems
Let be a support set and let be a fixed sign sequence on . In the setting of Theorem, consider the random vector … where is as in the source and the signs are not…
Let be the group of real orthogonal matrices, and let the least statistically-dependent basis (LSDB) be the basis obtained by optimizing statistical ind…
Strong lottery ticket conjecture at virtually all sparsity levels. There exists such that, with probability at least , the network contains a mask with sparsity …
Let be an -sparse mean vector in the horizontal-split distributed mean-estimation model, where each of machines holds a subset of independent sam…
The partition characterisation conjecture. The graph is -tight if and only if, for every edge , there exists a partition of such that
Deep neural networks are trained by gradient descent with small initialization; in this setting, the training dynamics and the notion of solution complexity are considered in the f…
Stronger edge bound conjecture. The number of edges satisfies
Dimensional sparsity-rate conjecture. In dimensions, the support of should shrink at rate
For a polynomial , let denote the number of matrices with characteristic polynomial . Characteristic-polynomial s…
Block-and-hole minimal-rigidity conjecture. The following statements are equivalent:
-sparsity conjecture. Every -tight simple graph is minimally rigid in .
Let ) be a graph, where denotes the -dimensional rigidity matroid and an -bridge is an edge whose deletion lowers the rank in that matroid.…
Let be the number of covariates and the sample size. In the regime , the paper's DCal estimator achieves -consistency under the minimal sparsity conditio…
The matrix is the quadratic coefficient matrix, and MILO denotes the mixed-integer linear optimization formulation obtained by reformulating the convex quadratic optimization p…
Thurstone's conjecture. Simple structure resolves the rotational invariance of factor analysis.
Let be an arbitrary integer vector, let , and consider the knapsack feasibility problem … Write for the number of…
N-1 redundancy sparsity conjecture. This sparsity is a natural property of the N-1 redundancy engineered into electrical networks.
Sparsity-promoting -norm conjecture. The norm may promote sparsity in ways similar to those in compressive sensing.
Optimality conjecture. The lower rate is optimal for all admissible configurations of , , and sparsity parameters.
Unknown-variance extension conjecture. The results of Theorem concerning the ABOS property of mBIC1, mBIC2, and mBIC3 should also hold when is unknown.
ABOS conjecture for mBIC1–mBIC3. In analogy with the known-variance case, mBIC1, mBIC2, and mBIC3 should be ABOS for a wide range of sparsity levels when is unknown.
Let be an operator acting on a basis , and consider the normalized pairwise values … The operator is said to act well on the basis when these values are not too large…