Posters

This paper proves fundamental space complexity lower bounds for approximating logistic loss, revealing that existing coreset constructions are surprisingly optimal.

This paper introduces a selective G-Bispectrum algorithm, slashing the computational complexity from O(|G|^2) to O(|G|), making G-invariant deep learning faster and more scalable.

Beat the 1/e barrier in the secretary problem using only an additive gap prediction!

Gradient descent’s sample complexity in non-smooth stochastic convex optimization is Õ(d/m+1/√m), matching worst-case ERMs and showing no advantage over naive methods.

LLMs use different internal structures for few-shot learning and fine-tuning, showing a transition in the middle network layers that impacts information encoding and task solving strategies.

OKRidge’s reliability for solving sparse ridge regression problems is rigorously proven through theoretical error analysis, enhancing its applicability in machine learning.

Optimization’s implicit bias in robust machine learning hurts generalization; this work reveals how algorithm/architecture choices impact robustness, suggesting better optimization strategies are need…

Neural networks exhibit ‘Neural Regression Collapse’ (NRC) during training, where feature vectors collapse to subspaces spanned by principal components of features and weights, and the weight vector G…

Hard attention transformers show surprisingly greater power when processing numerical data sequences, exceeding capabilities on string data; this advancement is theoretically analyzed via circuit comp…

Federated learning gets a speed boost: New extrapolation strategies significantly improve FedProx’s convergence, offering both theoretical backing and practical enhancements.

Poisson Midpoint Method quadratically accelerates Langevin Monte Carlo for diffusion models, achieving high-quality image generation with significantly fewer computations.

Neural circuits for goal-directed navigation are modeled as Lie group operator searches, implemented by a two-layer feedforward circuit mimicking Drosophila’s navigation system.

Researchers found the minimax optimal rate of HSIC estimation for translation-invariant kernels is O(n⁻¹/²), settling a two-decade-old open question and validating many existing HSIC estimators.

Neuromap leverages graph neural networks to optimize the map equation for community detection, achieving competitive performance and automatically determining the optimal number of clusters.

A new, surprisingly simple boosting algorithm achieves provably optimal sample complexity and outperforms existing algorithms on large datasets.

This research dramatically accelerates and improves hybrid language models by distilling large Transformers into linear RNNs, achieving performance comparable to the original Transformer with signific…

This paper presents computationally efficient transfer reinforcement learning algorithms that remove the dependence on state/action space sizes while achieving minimax optimality.

This paper reveals fundamental limits of differential privacy in online learning, demonstrating a clear separation between pure, approximate, and non-private settings.

Deep RL policy learning is improved by identifying and boosting key parameter update directions using a novel temporal SVD analysis, leading to more efficient and effective learning.

I-OBS, a novel family of sparse recovery algorithms leveraging second-order information, achieves faster convergence rates for sparse DNNs, validated by large-scale experiments.