arXiv research

PSMM method optimizes matrix sufficient dimension reduction.

A1GM method improves efficiency in reconstructing missing data using KL divergence.

This paper establishes risk convergence and asymptotic weight matrix alignment --- a form of implicit regularization --- of gradient flow and gradient descent when applied to deep linear networks on linearly separable data. In more detail, for gradient flow applied to strictly decreasing loss functions (with similar re…

Volume comparison theorem for rank 1 symmetric spaces proved.

Gradient Descent with small random initialization solves rank-1 matrix completion efficiently.

Paper presents a rank-1 approximation method for natural policy gradients in deep RL.

No non-product Hessian rank 1 affine homogeneous hypersurfaces exist in dimensions 5 and above.

Efficiently representing real world data in a succinct and parsimonious manner is of central importance in many fields. We present a generalized greedy pursuit framework, allowing us to efficiently solve structured matrix factorization problems, where the factors are allowed to be from arbitrary sets of structured vect…

If the group of a 2-knot group $K$ has an abelian normal subgroup of rank $\geq1$ which is not finitely generated then either $K$ has no minimal Seifert hypersurface or $K$ is topologically equivalent to Example 10 of Ralph Fox's``{\it A quick trip through knot theory}".

Rank-1 BNNs improve efficiency and scalability of Bayesian neural nets.

We explore the algebraic structure of the solution space of convex optimization problem Constrained Minimum Trace Factor Analysis (CMTFA), when the population covariance matrix $Σ_x$ has an additional latent graphical constraint, namely, a latent star topology. In particular, we have shown that CMTFA can have either a …

Abstract: Proves no non-trivial normal orbits for specific Hamiltonians.

Optimizes matching in weighted graphs with semi-bandit sampling.

Four algorithms improve sparse tensor BR1Approx with theoretical guarantees.

Algorithm recovers factors of rank-1 matrices from noisy measurements.

Consider a movie recommendation system where apart from the ratings information, side information such as user's age or movie's genre is also available. Unlike standard matrix completion, in this setting one should be able to predict inductively on new users/movies. In this paper, we study the problem of inductive matr…

For a manifold with nonpositive curvature, the Martin boundary is described by the behavior of normalized Green's functions at infinity. A classical result by Anderson and Schoen states that if the manifold has pinched negative curvature, the geometric boundary is the same as the Martin boundary. In this paper, we stud…

New method predicts and optimizes matrix recovery from noisy measurements.

Estimates rank-one spikes from heavy-tailed noise using self-avoiding walks.

In this paper we consider the Poisson algebraic structure associated with a classical $r$-matrix, i.e. with a solution of the modified classical Yang--Baxter equation. In Section 1 we recall the concept and basic facts of the $r$-matrix type Poisson orbits. Then we describe the $r$-matrix Poisson pencil (i.e the pair o…

We propose stochastic rank-$1$ bandits, a class of online learning problems where at each step a learning agent chooses a pair of row and column arms, and receives the product of their values as a reward. The main challenge of the problem is that the individual values of the row and column are unobserved. We assume tha…

Proves conditions for Fourier transforms in rank 1 symmetric spaces.

Solves the Wiegold problem by showing free products of left-orderable groups have normal rank > 1.

Study shows dynamics of rank 1 orbifolds in flat surfaces.

PSI-LinUCB improves scalability for large recommender systems.

We show that a simple randomized sketch of the matrix multiplicative weight (MMW) update enjoys (in expectation) the same regret bounds as MMW, up to a small constant factor. Unlike MMW, where every step requires full matrix exponentiation, our steps require only a single product of the form $e^A b$, which the Lanczos …

Sharp isoperimetric inequalities for Neumann eigenvalues in symmetric spaces.

Sparse PCA is a widely used technique for high-dimensional data analysis. In this paper, we propose a new method called low-rank principal eigenmatrix analysis. Different from sparse PCA, the dominant eigenvectors are allowed to be dense but are assumed to have a low-rank structure when matricized appropriately. Such a…

The probability that a user will click a search result depends both on its relevance and its position on the results page. The position based model explains this behavior by ascribing to every item an attraction probability, and to every position an examination probability. To be clicked, a result must be both attracti…

Right inverse found for Cartan differential in rank-1 symmetric spaces.

This expository monograph cuts a short path from the common, elementary background in geometry (linear algebra, vector bundles, and algebraic ideals) to the most advanced theorems about involutive exterior differential systems: (1) The incidence correspondence of the characteristic variety, (2) Guillemin normal form an…

The paper describes hyperkähler geometry of cotangent bundles using rank-1 projections.

GD learns matrix solutions incrementally, revealing insights into generalization.

This work is concerned with the non-negative rank-1 robust principal component analysis (RPCA), where the goal is to recover the dominant non-negative principal components of a data matrix precisely, where a number of measurements could be grossly corrupted with sparse and arbitrary large noise. Most of the known techn…

Simple AMP algorithm robust to adversarial corruption.

The class ${\sf IV}_2$ of $2$-nondegenerate constant Levi rank $1$ hypersurfaces $M^5 \subset \mathbb{C}^3$ is governed by Pocchiola's two primary invariants $W_0$ and $J_0$. Their vanishing characterizes equivalence of such a hypersurface $M^5$ to the tube $M_{\sf LC}^5$ over the real light cone in $\mathbb{R}^3$. Whe…

Study on diagonal and separating coordinates for symmetric spaces of rank 1.

Weight normalization speeds up matrix sensing problems.

We establish the proportionality principle between the Riemannian volume and locally finite simplicial volume for Q-rank 1 locally symmetric spaces covered by products of hyperbolic spaces, giving the first examples for manifolds whose cusp groups are not necessarily amenable. Also, we give a simple direct proof of the…

Layer normalization with activations prevents Gram matrix rank collapse at initialization.

Improves detection of low-rank signals from noisy data matrices.

We use generalised cross--ratios to prove the Ptolemaean inequality and the Theorem of Ptolemaeus in the setting of the boundary of symmetric Riemannian spaces of rank 1 and of negative curvature.

Based on a new atomic norm, we propose a new convex formulation for sparse matrix factorization problems in which the number of nonzero elements of the factors is assumed fixed and known. The formulation counts sparse PCA with multiple factors, subspace clustering and low-rank sparse bilinear regression as potential ap…

The Cayley hyperbolic space minimizes volume entropy among finite-volume metrics.

For a given lattice, we establish an equivalence involving a closed zone of the corresponding Voronoi polytope, a lamina hyperplane of the corresponding Delaunay partition and a quadratic form of rank 1 being an extreme ray of the corresponding L-type domain.

This paper develops the exact linear relationship between the leading eigenvector of the unnormalized modularity matrix and the eigenvectors of the adjacency matrix. We propose a method for approximating the leading eigenvector of the modularity matrix, and we derive the error of the approximation. There is also a comp…

We study rank $1$ flat bundles over solvmanifolds whose cohomologies are non-trivial. By using Hodge theoretical properties for all topologically trivial rank $1$ flat bundles, we represent the structure theorem of Kähler solvmanifolds as extensions of Hasegawa's result and Benson-Gordon's result for nilmanifolds.

Sign-RIP improves robust low-rank matrix recovery by preserving norms even with corrupted measurements.