AI Theory

Random Function Descent (RFD) replaces the classical convex function framework with a random function approach, providing a scalable gradient descent method with inherent scale invariance and a theore…

Random Cycle Coding (RCC) optimally compresses cluster assignments in large datasets, saving up to 70% storage in vector databases by eliminating the need for integer IDs.

QWO: a novel method dramatically speeds up permutation-based causal discovery in linear Gaussian models, enabling the analysis of larger datasets and advancing causal inference.

Novel algorithms stabilize multi-server queueing systems with unknown service rates, achieving sublinear regret by learning server preferences via preference feedback.

Novel algorithms for query-efficient correlation clustering with noisy oracles achieve a balance between query complexity and solution quality, offering theoretical guarantees and outperforming baseli…

Quantum algorithms achieve speedups in non-smooth, non-convex optimization, outperforming classical methods by a factor of ε⁻²/³ in query complexity for finding (δ,ε)-Goldstein stationary points.

Quantum speedup achieved for large-scale market equilibrium computation!

New orthogonal learner quantifies treatment effect’s randomness, providing sharper insights beyond average effects.

Researchers introduce QIM-compatibility, a novel framework for modeling qualitative relationships in probability distributions using directed hypergraphs, significantly expanding beyond standard condi…

Q2VMC, a novel quantum chemistry algorithm, drastically boosts the efficiency and accuracy of solving the Schrödinger equation using a quadratic update mechanism and neural network ansatzes.

Researchers improve two-sided matching market algorithms by prioritizing stability through novel bandit-learning algorithms, providing theoretical bounds on sample complexity and demonstrating intrigu…

Leveraging public data enhances differentially private (DP) learning, but its limits are unclear. This paper establishes tight theoretical bounds for DP stochastic convex optimization, revealing when …

Pseudo-Private Data Guided Model Inversion (PPDG-MI) significantly improves model inversion attacks by dynamically tuning the generative model to increase the sampling probability of actual private da…

Unmeasured confounders hinder causal inference; this paper introduces a novel method using two pre-treatment text instances and zero-shot models to infer proxies for unobserved confounders, enabling p…

POETRY: a recursive neural theorem prover achieving 5.1% higher success rate and solving substantially longer proofs.

Verifiably safe AI controllers are created via a novel framework, VerSAILLE, which uses differential dynamic logic and open-loop NN verification to prove safety for unbounded time horizons.

Researchers achieve provably optimal memory capacity in transformer-compatible Hopfield models by framing the problem as an optimal spherical code arrangement, resulting in a novel sublinear time algo…

Deep learning’s generalization ability defies conventional wisdom; this paper proves that overfitting in deep neural networks is ’tempered’, neither catastrophic nor perfectly benign, for both minimal…

PREPARED efficiently edits DNNs to provably satisfy properties by relaxing the problem to a linear program, minimizing parameter changes.

Complex numbers boost neural network performance! This study proves that complex parameterizations in structured state space models (SSMs) enable more efficient and practical learning of complex mappi…