arXiv research

We give a reduction from {\sc clique} to establish that sparse PCA is NP-hard. The reduction has a gap which we use to exclude an FPTAS for sparse PCA (unless P=NP). Under weaker complexity assumptions, we also exclude polynomial constant-factor approximation algorithms.

Detecting adversarial examples is as hard as classifying them.

New proof shows a link problem is hard without complex links.

This paper explores the computational hardness of generating latent vectors for generative models.

Introduces a continuous version of LWE problem.

Reduces learning periodic neural networks to lattice problems, proving hardness under cryptographic assumptions.

Computing PL geometric category in 2D is NP-hard.

We show that {\sc Heegaard Genus $\leq g$}, the problem of deciding whether a triangulated 3-manifold admits a Heegaard splitting of genus less than or equal to $g$, is NP-hard. The result follows from a quadratic time reduction of the NP-complete problem {\sc CNF-SAT} to {\sc Heegaard Genus $\leq g$}.

We prove that for every $d\geq 2$, deciding if a pure, $d$-dimensional, simplicial complex is shellable is NP-hard, hence NP-complete. This resolves a question raised, e.g., by Danaraj and Klee in 1978. Our reduction also yields that for every $d \ge 2$ and $k \ge 0$, deciding if a pure, $d$-dimensional, simplicial com…

Hard to estimate $L^2$-accurate scores without strong assumptions.

New insights link diverse statistical problems via secret leakage planted clique.

Adapts DR objectives for both sample and feature size reduction.

SLR tackles sparse linear regression problems, showing hardness for efficient algorithms.

Hard visual attention is a promising approach to reduce the computational burden of modern computer vision methodologies. Hard attention mechanisms are typically non-differentiable. They can be trained with reinforcement learning but the high-variance training this entails hinders more widespread application. We show h…

New lower bounds for linear classification problems in high dimensions.

In the past decade, sparse principal component analysis has emerged as an archetypal problem for illustrating statistical-computational tradeoffs. This trend has largely been driven by a line of research aiming to characterize the average-case complexity of sparse PCA through reductions from the planted clique (PC) con…

We prove that the problem of deciding whether a 2- or 3-dimensional simplicial complex embeds into $\mathbb{R}^3$ is NP-hard. Our construction also shows that deciding whether a 3-manifold with boundary tori admits an $\mathbb{S}^{3}$ filling is NP-hard. The former stands in contrast with the lower dimensional cases wh…

Paper identifies reductive MDPs, solving them in polynomial time.

TQFT invariants are either easy or hard to compute, depending on the TQFT type.

The application of stochastic variance reduction to optimization has shown remarkable recent theoretical and practical success. The applicability of these techniques to the hard non-convex optimization problems encountered during training of modern deep neural networks is an open problem. We show that naive application…

Attention operators have been widely applied in various fields, including computer vision, natural language processing, and network embedding learning. Attention operators on graph data enables learnable weights when aggregating information from neighboring nodes. However, graph attention operators (GAOs) consume exces…

Topological quantum computers use hyperbolic knots for computations.

FibeRed reduces complex data dimensions while preserving topology.

Detects dense subhypergraphs in random hypergraphs using low-degree polynomials.

Equity-Transformer solves NP-hard min-max routing problems efficiently.

While on some natural distributions, neural-networks are trained efficiently using gradient-based algorithms, it is known that learning them is computationally hard in the worst-case. To separate hard from easy to learn distributions, we observe the property of local correlation: correlation between local patterns of t…

DM uses semigroup property to tune diffusion time for better data analysis.

We propose an efficient protocol for decentralized training of deep neural networks from distributed data sources. The proposed protocol allows to handle different phases of model training equally well and to quickly adapt to concept drifts. This leads to a reduction of communication by an order of magnitude compared t…

SapAugment learns adaptive augmentation policies for better model training.

We focus in this paper on dataset reduction techniques for use in k-nearest neighbor classification. In such a context, feature and prototype selections have always been independently treated by the standard storage reduction algorithms. While this certifying is theoretically justified by the fact that each subproblem …

Paper proposes a novel unsupervised feature selection method using K-means and ADMM.

Two-layer networks learn hard GLMs with SGD in high dimensions.

Paper develops KMS Wasserstein for high-dimensional data reduction.

Efficiently transforms Gaussian data to simulate various target distributions.

Study bridges welfare maximization and CATE estimation in policy learning.

Study shows exponential error reduction in multiclass classification without bias-variance trade-off.

Normal surface theory, a tool to represent surfaces in a triangulated 3-manifold combinatorially, is ubiquitous in computational 3-manifold theory. In this paper, we investigate a relaxed notion of normal surfaces where we remove the quadrilateral conditions. This yields normal surfaces that are no longer embedded. We …

Stochastic convex optimization algorithms are the most popular way to train machine learning models on large-scale data. Scaling up the training process of these models is crucial, but the most popular algorithm, Stochastic Gradient Descent (SGD), is a serial method that is surprisingly hard to parallelize. In this pap…

We consider the problem of computing the best-fitting ReLU with respect to square-loss on a training set when the examples have been drawn according to a spherical Gaussian distribution (the labels can be arbitrary). Let $\mathsf{opt} < 1$ be the population loss of the best-fitting ReLU. We prove: 1. Finding a ReLU wit…

We consider robust optimization problems, where the goal is to optimize in the worst case over a class of objective functions. We develop a reduction from robust improper optimization to Bayesian optimization: given an oracle that returns $α$-approximate solutions for distributions over objectives, we compute a distrib…

New lower bound shows RL with linear approximations is computationally hard.

Let EMBED(k,d) be the following algorithmic problem: Given a finite simplicial complex K of dimension at most k, does there exist a (piecewise linear) embedding of K into R^d? Known results easily imply polynomiality of EMBED(k,2) (k=1,2; the case k=1, d=2 is graph planarity) and of EMBED(k,2k) for all k>2 (even if k i…

In this paper we consider the dictionary learning problem for sparse representation. We first show that this problem is NP-hard by polynomial time reduction of the densest cut problem. Then, using successive convex approximation strategies, we propose efficient dictionary learning schemes to solve several practical for…

Unsupervised dimension selection is an important problem that seeks to reduce dimensionality of data, while preserving the most useful characteristics. While dimensionality reduction is commonly utilized to construct low-dimensional embeddings, they produce feature spaces that are hard to interpret. Further, in applica…

We describe a new variational lower-bound on the minimum energy configuration of a planar binary Markov Random Field (MRF). Our method is based on adding auxiliary nodes to every face of a planar embedding of the graph in order to capture the effect of unary potentials. A ground state of the resulting approximation can…

Paper proves hardness of learning various complex models under local pseudorandom generators.

Online algorithms stabilize in feedback loops of performative prediction.

Proves SQ lower bounds for learning two-hidden-layer neural networks.