Deep Learning

ERM’s efficiency in feature learning surprisingly remains high even with massive feature maps; its excess risk asymptotically matches an oracle procedure’s, implying potential for streamlined feature-…

Learning sparse functions efficiently with gradient methods is challenging; this paper introduces Differentiable Learning Queries (DLQ) to precisely characterize gradient query complexity, revealing s…

This paper introduces novel autoregressive sampling and hybrid training strategies for score-based diffusion models, significantly boosting PDE forecasting and assimilation accuracy.

This paper presents a novel, efficient method for Jacobian nuclear norm regularization in deep learning, replacing computationally expensive SVDs with equivalent Frobenius norm computations, thereby e…

Nonstationary Sparse Spectral Permanental Process (NSSPP) enhances point process modeling by using sparse spectral representations, enabling flexible, efficient, nonstationary kernel learning.

Adaptive algorithm selection for noisy ICA is achieved via a novel nonparametric independence score, improving accuracy and efficiency.

Overparameterized ConvResNets surprisingly excel at prediction; this study proves they efficiently learn smooth functions on low-dimensional manifolds, avoiding the curse of dimensionality.

New non-parametric classifiers using expand-and-sparsify representation achieve minimax-optimal convergence, adapting to low-dimensional manifold structure.

Noether’s Razor learns conserved quantities and symmetries directly from data via Bayesian model selection, improving dynamical systems modeling accuracy and generalizability.

Neuronal Competition Groups (NCGs) enhance supervised STDP training in spiking neural networks by promoting balanced competition and improved class separation, resulting in significantly higher classi…

NeuralFuse: A novel add-on module learns input transformations to maintain accuracy in low-voltage DNN inference, achieving up to 57% accuracy recovery and 24% energy savings without retraining.

Neural P³M enhances geometric GNNs by incorporating mesh points to model long-range interactions in molecules, achieving state-of-the-art accuracy in predicting energy and forces.

Neural Flow Diffusion Models (NFDM) revolutionize generative modeling by introducing a learnable forward process, resulting in state-of-the-art likelihoods and versatile generative dynamics.

Neural Embeddings Rank (NER) aligns 3D latent neural dynamics with movements, enabling cross-session decoding and revealing consistent neural dynamics across brain areas.

Neural Conditional Probability (NCP) offers a new operator-theoretic approach for efficiently learning conditional distributions, enabling streamlined inference and providing theoretical guarantees fo…

Researchers enhance imbalanced data classification by inducing Neural Collapse to Multiple Centers (NCMC) using a novel cosine loss function, achieving performance comparable to state-of-the-art metho…

Neural Collapse-inspired Feature Alignment (NCFAL) significantly boosts out-of-distribution generalization by aligning semantic features to a simplex ETF, even without environment labels.

Researchers introduce Geometric Parameterization (GmP), a novel neural network parameterization resolving instability in ReLU network training, leading to faster convergence and better generalization.

ChemFlow: a new framework efficiently explores chemical space using latent flows, unifying existing methods & incorporating physical priors for molecule manipulation and optimization.

f-DIME: a novel class of discriminative mutual information estimators using f-divergence outperforms state-of-the-art methods by achieving an excellent bias-variance trade-off. This is achieved throug…