Gradient descent with polylogarithmic width achieves arbitrarily low test error for shallow ReLU networks.
problem Achieving low test error with shallow ReLU networks using gradient descent.
method Gradient descent with polylogarithmic width and polylogarithmic number of samples.
result Gradient descent achieves arbitrarily low test error with shallow ReLU networks of polylogarithmic width.
Study shows sample complexity for multicalibration is Θ(ε^-3) with polylogarithmic factors.
problem Minimizing Expected Calibration Error (ECE) for predictors with respect to a family of groups.
method Proved necessary and sufficient sample complexity of Θ(ε^-3) for multicalibration, using online-to-batch reduction and lower bounds.
result Sample complexity of multicalibration is Θ(ε^-3) with polylogarithmic factors, distinguishing it from marginal calibration.
New method uses higher-order Langevin dynamics for efficient parallel sampling.
problem Efficient parallel sampling from high-dimensional log-concave distributions.
method Combines higher-order Langevin dynamics with blockwise Lagrange polynomial interpolation.
result Reduces the number of parallel points required for a target accuracy.
Paper solves no-swap regret minimization for combinatorial bandits with polylogarithmic dependence on N.
problem Design efficient no-swap regret algorithms for combinatorial bandits with exponentially large action space.
method Introduces a no-swap-regret learning algorithm with polylogarithmic dependence on N and demonstrates efficient implementation.
result Achieves no-swap regret with polylogarithmic dependence on N, resolving an open problem.
New algorithm selects best distribution privately in nearly-linear time.
problem Estimating the best distribution from samples under differential privacy constraints.
method Differentially private algorithm with nearly-linear time complexity and optimal approximation factor.
result Achieves optimal approximation factor of 3 with modest sample complexity increase.
Sampling logconcave functions arising in statistics and machine learning has been a subject of intensive study. Recent developments include analyses for Langevin dynamics and Hamiltonian Monte Carlo (HMC). While both approaches have dimension-independent bounds for the underlying c o n t i n u o u s \mathit{continuous} continuous processes under s…
Quantum machine learning can't achieve polylogarithmic runtimes, even with quantum data access.
problem Bounding the minimum number of samples required for supervised quantum learning.
method Statistical learning theory and quantum machine learning algorithms.
result Quantum machine learning algorithms for supervised learning have at most polynomial speedups over classical algorithms.
Neural networks can achieve optimal sample complexity for learning single-index models.
problem Achieving optimal computational-statistical tradeoff in learning Gaussian single-index models.
method Unified gradient-based algorithm for training a two-layer neural network, adaptable to various loss and activation functions.
result Sample complexity of d s ⋆ / 2 ∨ d d^{s^\star/2} \lor d d s ⋆ /2 ∨ d matches the SQ lower bound up to a polylogarithmic factor. Mirzakhani volumes of moduli spaces are polylogarithmic.
problem Understanding the volume of moduli spaces of hyperbolic surfaces.
method Expressed as a sum of polylogarithms evaluated at specific points.
result Mirzakhani volumes are polylogarithmic.
Study higher genus polylogarithms under Riemann surface degenerations.
problem Understanding higher genus polylogarithms under degenerations.
method Investigate the Enriquez connection for polylogarithms and show it becomes a known connection for families of Riemann surfaces.
result Higher genus polylogarithms can be described explicitly as power series in deformation parameters and logarithms of families.
Investigates webs related to cluster algebras and polylogarithms.
problem Understanding webs associated with cluster algebras and polylogarithms.
method Introducing AMP webs and analyzing their properties, proving results and conjectures.
result Many webs associated with polylogarithms and cluster algebras are AMP webs.
Efficiently estimates binary product distributions with privacy.
problem Estimating means of binary product distributions privately and accurately.
method Polynomial time, pure differential privacy approach.
result Optimal sample complexity with polylogarithmic factors.
QATS efficiently decodes HMMs with polylogarithmic complexity.
problem Efficiently decoding hidden Markov models from noisy observations.
method Divide-and-conquer procedure with polylogarithmic sequence complexity and cubic state space complexity.
result QATS outperforms Viterbi and PMAP in speed and accuracy.
Private optimization faster on interpolation problems with quadratic growth.
problem Private optimization in interpolation problems.
method Adaptive algorithm with improved sample complexity.
result Exponential improvement in private sample complexity for quadratic growth.
A new algorithm estimates mean adaptively to covariance, faster and more flexible than existing methods.
problem Estimating mean of a distribution with unknown covariance efficiently and privately.
method Adaptive differentially private algorithm with optimal convergence rates and near-linear sample complexity.
result Achieves optimal rates of convergence with respect to the Mahalanobis norm ∣ ∣ ⋅ ∣ ∣ Σ ||\cdot||_Σ ∣∣ ⋅ ∣ ∣ Σ . In this paper, we settle the sampling complexity of solving discounted two-player turn-based zero-sum stochastic games up to polylogarithmic factors. Given a stochastic game with discount factor γ ∈ ( 0 , 1 ) γ\in(0,1) γ ∈ ( 0 , 1 ) we provide an algorithm that computes an ε ε ε -optimal strategy with high-probability given $\tilde{O}((1 - γ)^{-3}…
New algorithms learn from untrusted batches with improved efficiency.
problem Learning from untrusted batches with adversarial responses.
method Sum-of-Squares hierarchy applied to robust mean estimation.
result Reduces sample complexity to polylogarithmic in n n n for most natural distributions. We present an efficient and practical algorithm for the online prediction of discrete-time linear dynamical systems with a symmetric transition matrix. We circumvent the non-convex optimization problem using improper learning: carefully overparameterize the class of LDSs by a polylogarithmic factor, in exchange for con…
New research shows deep ReLU networks can be learned with polylogarithmic width.
problem Learning deep ReLU networks with limited over-parameterization.
method Using gradient descent, the study establishes learning guarantees for networks with polylogarithmic width.
result Deep ReLU networks can be learned with a polylogarithmic width condition, not just a high degree polynomial.
Standard results in stochastic convex optimization bound the number of samples that an algorithm needs to generate a point with small function value in expectation. More nuanced high probability guarantees are rare, and typically either rely on "light-tail" noise assumptions or exhibit worse sample complexity. In this …
A wide range of fundamental machine learning tasks that are addressed by the maximum a posteriori estimation can be reduced to a general minimum conical hull problem. The best-known solution to tackle general minimum conical hull problems is the divide-and-conquer anchoring learning scheme (DCA), whose runtime complexi…
New algorithm reduces regret in online portfolio and quantum state learning.
problem Efficiently learning portfolios and quantum states online with minimal regret.
method BISONS algorithm for online portfolio selection, SCHRODINGER'S BISONS for quantum states, with polylogarithmic regret.
result First efficient algorithm with polylogarithmic regret for online portfolio selection and quantum states.
Improved GNN simulation of WL test with exponentially lower complexity.
problem Improving the complexity of simulating the Weisfeiler-Lehman test with GNNs.
method Exponentially lower complexity simulation of WL test using GNNs with polylogarithmic parameters and O(log n) bits feature vectors.
result Near-optimal construction with logarithmic lower bounds for feature vector length and neural network size.
Piecewise polynomial interpolation-based gradient descent reduces oracle complexity for smooth loss functions.
problem Optimizing empirical risk minimization loss functions
method Piecewise polynomial interpolation-based gradient descent
result Oracle complexity is reduced for smooth loss functions
New method proves fast regret bounds for online RLHF with generalized preferences.
problem Minimizing max-regret in online RLHF with general preferences and bandit feedback.
method Adopted Generalized Bilinear Preference Model (GBPM) to investigate polylogarithmic regret guarantees.
result Proved polylogarithmic regret bounds for Greedy Sampling and Explore-Then-Commit policies under GBPM.
Optimal sampling bounds for various classification losses under different regularization terms.
problem Achieving optimal sampling complexity for classification losses under different regularization terms.
method Proved optimal sampling bounds for a broad class of Lipschitz continuous classification loss functions under various regularization terms.
result Proved k 2 / ε 2 k^2/\varepsilon^2 k 2 / ε 2 upper and lower bounds for ∥ ⋅ ∥ 2 / k \|\cdot\|_2/k ∥ ⋅ ∥ 2 / k regularization, and k / ε 2 k/\varepsilon^2 k / ε 2 upper and lower bounds for ∥ ⋅ ∥ 1 / k \|\cdot\|_1/k ∥ ⋅ ∥ 1 / k regularization. We prove optimal subspace embedding conjecture up to sub-polylogarithmic factors.
problem Optimal dimension and sparsity of subspace embeddings.
method Iterative decoupling technique to analyze higher-order trace moment bounds.
result Sub-polylogarithmic factors in dimension and sparsity of subspace embeddings.
The paper explores the geometry of the Spence-Kummer trilogarithm equation and its Galois analogue.
problem Investigating the geometry and functional equation of the Spence-Kummer trilogarithm.
method Using algebraic relations between polylogarithm generating series and path systems, along with tensor and homotopy criteria for functional equations.
result Derives a precise form of the Spence-Kummer equation and its Galois analogue.
Neural network learns low-dimensional polynomials with SGD near information-theoretic limit.
problem Learning a single-index target function with gradient descent.
method Two-layer neural network optimized by SGD on squared loss.
result Sample and runtime complexity of n ≃ T = Θ ( d ⋅ p o l y l o g d ) n \simeq T = Θ(d\!\cdot\! \mathrm{polylog} d) n ≃ T = Θ ( d ⋅ polylog d ) for polynomial single-index models, matching information theoretic limit up to polylogarithmic factors. Motivated by a sampling problem basic to computational statistical inference, we develop a nearly optimal algorithm for a fundamental problem in spectral graph theory and numerical analysis. Given an n × n n\times n n × n SDDM matrix M {\bf \mathbf{M}} M , and a constant − 1 ≤ p ≤ 1 -1 \leq p \leq 1 − 1 ≤ p ≤ 1 , our algorithm gives efficient access to a…
Improved sampling from high-dimensional Gaussians using smoothed scores.
problem Sampling from high-dimensional Gaussian distributions with gradient information.
method Using smoothed scores, which are gradients of the logarithms of Gaussian-convolved densities, to overcome approximation barriers.
result Improved sampling efficiency with a complexity of \(O\left(\left(\logκ+\log(e\sqrt d/δ_{
m TV})
ight)\log(e\sqrt d/δ_{
m TV})
ight)\) smoothed-score queries.
New algorithm for RL with horizon-free reward-free exploration for linear MDPs.
problem Reward-free reinforcement learning with long planning horizons.
method Uncertainty-weighted value-targeted regression with exploration-driven pseudo-reward and moment estimator.
result Horizon-free sample complexity of O ( d 2 ε − 2 ) O(d^2\varepsilon^{-2}) O ( d 2 ε − 2 ) for finding an ε \varepsilon ε -optimal policy. Paper presents a faster classical algorithm for principal component regression.
problem Efficiently solving principal component regression problems.
method Uses quantum-inspired linear algebra techniques.
result Achieves polylogarithmic runtime, significantly faster than state-of-the-art.
Protocol learns pure quantum states with minimal disturbance.
problem Efficiently learn quantum states with minimal disturbance.
method Sequential measurements with minimal disturbance.
result Achieves maximal precision with polylogarithmic regret.
We present a novel parallelisation scheme that simplifies the adaptation of learning algorithms to growing amounts of data as well as growing needs for accurate and confident predictions in critical applications. In contrast to other parallelisation techniques, it can be applied to a broad class of learning algorithms …
Efficiently estimates Gaussian distributions privately and robustly.
problem Private and robust estimation of Gaussian distributions.
method Efficient algorithms for pure and approximate differential privacy models.
result Optimal sample complexity in both pure and approximate differential privacy settings.
Kähler information manifolds for signal filters in weighted Hardy spaces are explored.
problem Developing a geometric framework for signal processing filters in weighted Hardy spaces.
method Introducing weighted Hardy spaces and smooth transformations of transfer functions, demonstrating the Kähler manifold structure.
result The Riemannian geometry of weighted Hardy norms for transfer functions forms a Kähler manifold.
Paper analyzes online tensorial ICA convergence with stochastic approximation.
problem Online tensorial ICA convergence analysis.
method Stochastic approximation for nonconvex optimization.
result Sharp finite-sample error bound of O ~ ( d / T ) \tilde{O}(\sqrt{d/T}) O ~ ( d / T ) . New bounds show current methods overestimate system parameter errors.
problem Current bounds overestimate parameter errors in system identification.
method Utilized asymptotic normality and second-order decomposition.
result Obtained finite-sample bounds matching optimal rates up to constants.
Constructs manifolds from quantum codes with novel geometric properties.
problem Creating manifolds with specific geometric constraints.
method Reverse engineering manifolds from quantum code chain complexes.
result First examples of power law Z 2 \mathbb{Z}_2 Z 2 systolic freedom. Sparse covariance estimation in the vertical-split model achieves exponential improvement over dense estimates.
problem Minimax estimation error for distributed covariance matrix estimation in the vertical-split setting.
method Elementwise s s s -sparsity is shown to reduce communication and sample complexity. result Minimax lower bounds for 1 1 1 -sparse cross-covariance estimation are established. Quantum algorithm speeds up Lasso regression by quadratically faster per iteration.
problem Efficiently solving high-dimensional linear regression with L1-penalty.
method Pathwise LARS algorithm adapted for quantum computing, using minimum-finding subroutines.
result Quadratic speedup in computation time for both number of features and observations.
New study reveals a polynomial penalty for adapting to unknown margin parameters in batched nonparametric bandits.
problem Adapting to an unknown margin parameter in batched nonparametric bandits.
method Introduces the regret inflation criterion and develops RoBIN algorithm to achieve optimal regret inflation.
result The optimal regret inflation grows polynomially with the horizon T, characterized by a convex optimization problem.
The paper analyzes GD for KANs, deriving bounds for training, generalization, and privacy.
problem Training dynamics, generalization, and privacy properties of KANs.
method Gradient Descent (GD) analysis for two-layer KANs under logistic loss and NTK-separable assumption.
result Polylogarithmic width suffices for GD to achieve optimization and generalization rates under DP.
Improved guarantees for misspecified kernelized bandit optimization.
problem Misspecification in kernelized bandit optimization.
method Localization and domain splitting techniques.
result Logarithmic or polylogarithmic growth of misspecification amplification.
The paper proposes a method to estimate tensor regression parameters using low-rank and sparse Tucker decompositions.
problem Estimating tensor regression parameters from limited data.
method Low-rank and sparse Tucker decompositions, non-convex optimization, projected gradient descent.
result The method can linearly converge to an appropriate solution under certain conditions.
Privacy improves robustness in statistical estimation.
problem Sparse mean estimation under privacy constraints.
method Sum-of-Squares method and exponential-time mechanisms.
result Private algorithms matching optimal tradeoffs are not known, but achieved via Sum-of-Squares.
New algorithms for privately learning decision lists and halfspaces.
problem Private learning of decision lists and halfspaces.
method Differentially private algorithms for PAC and online models.
result Private algorithms match or surpass non-private guarantees.