archive
Every paper Pith has read. Search by title, abstract, or pith.
2685 papers in stat.ML · page 17
-
Kernel embedding solves continuum optimal transport without meshes
Continuum-marginal optimal transport: a mesh-free kernel method
-
Zero-drift fields identify measures for companion-elliptic kernels
Identifiability and Stability of Generative Drifting with Companion-Elliptic Kernel Families
-
Drifting field vanishes exactly when measures coincide
Identifiability and Stability of Generative Drifting with Companion-Elliptic Kernel Families
-
Divergence weights boost model averaging in small samples
A Divergence-Based Method for Weighting and Averaging Model Predictions
-
DecompKAN ties or beats baselines on 15 of 32 forecasting benchmark cases
DecompKAN: Decomposed Patch-KAN for Long-Term Time Series Forecasting
-
Score identity calibrates reduced stochastic models from lagged pairs
Conditional Score-Based Modeling of Effective Langevin Dynamics
-
Sliced OT reference beats entropy regularization for closer transport plans
Sliced-Regularized Optimal Transport
-
Sliced OT prior sharpens regularized transport plans
Sliced-Regularized Optimal Transport
-
New algorithm selects subdata approaching information optimum
Nearly Optimal Subdata Selection
-
Gromov-Wasserstein aligns distance matrices into consensus embeddings
Gromov-Wasserstein Methods for Multi-View Relational Embedding and Clustering
-
Synthetic data preserves predictions but warps treatment effects
Generative Synthetic Data for Causal Inference: Pitfalls, Remedies, and Opportunities
-
Generative models keep predictions accurate but distort causal effects
Generative Synthetic Data for Causal Inference: Pitfalls, Remedies, and Opportunities
-
Brain activity patterns inverted to recover stimulus emotions
Inverting Foundation Models of Brain Function with Simulation-Based Inference
-
Sufficient derivative changes recover full latent causal graphs
Causal Representation Learning from General Environments under Nonparametric Mixing
-
Markov blanket partition enables general domain adaptation
A General Representation-Based Approach to Multi-Source Domain Adaptation
-
Coupled bootstrap corrects bias from ML labels in regressions
Bootstrapping with AI/ML-generated labels
-
Residual connections block rank collapse in Transformers
Rank, Head-Channel Non-Identifiability, and Symmetry Breaking: A Precise Analysis of Representational Collapse in Transformers
-
PathBoost matches or beats GNNs on half of graph benchmarks
Path-Based Gradient Boosting for Graph-Level Prediction
-
Fermat distance powers minimax optimal semi-supervised classifier
High-dimensional Semi-supervised Classification via the Fermat Distance
-
Distillation cuts memorization transferred to student models
On the Memorization of Consistency Distillation for Diffusion Models
-
Markov network equals Bayesian posterior for structural states
Probabilistic Graphical Model using Graph Neural Networks for Bayesian Inversion of Discrete Structural Component States
-
Preconditioner must stabilize faster than t^-(α+1)/2 to keep CLT
When Does Dynamic Preconditioning Preserve the Polyak-Ruppert CLT? A Stabilization Threshold
-
Refugee matching benefits hold across evaluation methods
Robustness of Refugee-Matching Gains to Off-Policy Evaluation Choices
-
Anchoring at posterior mean speeds variational inference for sequential models with random
Anchored Variational Inference for Personalized Sequential Latent-State Models
-
Online Newton with accelerated sketching converges globally
Inference of Online Newton Methods with Nesterov's Accelerated Sketching
-
Adaptive rotations make MCMC samplers work on any structural model
MCMC with Adaptive Principal-Component Transformation: Rotation-Invariant Universal Samplers for Bayesian Structural System Identification
-
Spatio-channel clustering cuts CNN FLOPs by 81% in MRI analysis
Hierarchical Spatio-Channel Clustering for Efficient Model Compression in Medical Image Analysis
-
Memory Sinkhorn solves mismatched Schrödinger bridge
Nonlinear Non-Gaussian Density Steering with Input and Noise Channel Mismatch: Sinkhorn with Memory for Solving the Control-affine Schr\"{o}dinger Bridge Problem
-
Conditioned diffusion supplies evolving data for multi-agent coordination
CODA: Coordination via On-Policy Diffusion for Multi-Agent Offline Reinforcement Learning
-
ReLU networks approximate polynomials with dimension-free errors
Explicit integral representations and quantitative bounds for two-layer ReLU networks
-
ReLU networks represent polynomials with dimension-free L2 bounds
Explicit integral representations and quantitative bounds for two-layer ReLU networks
-
Spectral algorithms split into three risk regimes in high dimensions
Learning Curves and Benign Overfitting of Spectral Algorithms in Large Dimensions
-
O(n) random numbers condition any matrix to O(n)
Well-Conditioned Oblivious Perturbations in Linear Space
-
One-way attention blocks outcome leakage in causal transformers
MOCA: A Transformer-based Modular Causal Inference Framework with One-way Cross-attention and Cutting Feedback
-
ProEval needs 8-65x fewer samples for accurate AI model evaluation
ProEval: Proactive Failure Discovery and Efficient Performance Estimation for Generative AI Evaluation
-
Mixture model draws non-linear cluster boundaries without labels
Turtle shell clustering: A mixture approach to discriminative clustering with applications to flow cytometry and other data
-
Second-order unlearning retains geometric memory in optimizer state
Shape of Memory: a Geometric Analysis of Machine Unlearning in Second-Order Optimizers
-
Support penalty selects reliable two-stage recommender policies
CASP: Support-Aware Offline Policy Selection for Two-Stage Recommender Systems
-
Lengthscale scaling prevents degeneration in high-dim trust region BO
Rethinking Trust Region Bayesian Optimization in High Dimensions
-
Running score averages enable low-variance non-Markovian sampling
Score-Repellent Monte Carlo: Toward Efficient Non-Markovian Sampler with Constant Memory in General State Spaces
-
Variational autoencoder improves long-term customer revenue forecasts
CLVAE: A Variational Autoencoder for Long-Term Customer Revenue Forecasting
-
Spectral estimator drives mixed membership error to zero
Mixed Membership sub-Gaussian Models
-
WassersteinGrad clarifies AI weather forecasts by geometric averaging
Explanation of Dynamic Physical Field Predictions using WassersteinGrad: Application to Autoregressive Weather Forecasting
-
Design infers object type from physical data and requirements
Design, Cups, and Blankets. A Free-Energy-Principle-Based Approach to Product Design
-
Manifold projections enable geometry-aware federated SPDnet
FedSPDnet: Geometry-Aware Federated Deep Learning with SPDnet
-
Super Learner gains finite-sample coverage via conformal prediction
Conformalized Super Learner
-
SQUEAK keeps kernel storage near effective dimension size
Pack only the essentials: Adaptive dictionary learning for kernel ridge regression
-
Kernel estimator adapts proposal to cut rejection sampling waste
Pliable rejection sampling
-
SOC-ICNNs expand convex networks from polyhedral to smooth functions
SOC-ICNN: From Polyhedral to Conic Geometry for Learning Convex Surrogate Functions
-
Bernstein polynomials estimate any valid isotropic covariance in infinite dimensions
Nonparametric Estimation of Isotropic Covariance Function