theoretical analysis
In AI, theoretical analysis refers to the study of algorithms and models based on mathematical frameworks to understand their limits, convergence properties, and performance guarantees. This involves analyzing their behavior in various scenarios, including worst-case and average-case situations, often leading to insights into the complexity and robustness of AI systems.
- $\text{G}^2\text{M}$: A Generalized Gaussian Mirror Method to Boost Feature Selection Power
- A Unified Framework for Fair Graph Generation: Theoretical Guarantees and Empirical Advances
- A Unified Framework for the Transportability of Population-Level Causal Measures
- A learnability analysis on neuro-symbolic learning
- Additive Models Explained: A Computational Complexity Approach
- Adv-SSL: Adversarial Self-Supervised Representation Learning with Theoretical Guarantees
- Adversarial generalization of unfolding (model-based) networks
- Agnostic Active Learning Is Always Better Than Passive Learning
- Agnostic Active Learning Is Always Better Than Passive Learning
- An Analysis of Concept Bottleneck Models: Measuring, Understanding, and Mitigating the Impact of Noisy Annotations
- Analytic Energy-Guided Policy Optimization for Offline Reinforcement Learning
- Asymmetric REINFORCE for off-Policy Reinforcement Learning: Balancing positive and negative rewards
- Attention Sinks: A 'Catch, Tag, Release' Mechanism for Embeddings
- Attention-based clustering
- Balancing Positive and Negative Classification Error Rates in Positive-Unlabeled Learning
- Blindfolded Experts Generalize Better: Insights from Robotic Manipulation and Videogames
- BlockScan: Detecting Anomalies in Blockchain Transactions
- CCL: Causal-aware In-context Learning for Out-of-Distribution Generalization
- CLIPTTA: Robust Contrastive Vision-Language Test-Time Adaptation
- Co-Regularization Enhances Knowledge Transfer in High Dimensions
- Combinatorial Ski Rental Problem: Robust and Learning-Augmented Algorithms
- Competitive Advantage Attacks to Decentralized Federated Learning
- Contrastive Learning with Data Misalignment: Feature Purity, Training Dynamics and Theoretical Generalization Guarantees
- Counteractive RL: Rethinking Core Principles for Efficient and Scalable Deep Reinforcement Learning
- Diffusion Adaptive Text Embedding for Text-to-Image Diffusion Models
- Distributed mediation analysis with communication efficiency
- Dynamic Bundling with Large Language Models for Zero-Shot Inference on Text-Attributed Graphs
- Efficient Representativeness-Aware Coreset Selection
- Enhancing Contrastive Learning with Variable Similarity
- Enhancing Deep Batch Active Learning for Regression with Imperfect Data Guided Selection
- Enhancing Graph Classification Robustness with Singular Pooling
- Error Feedback under $(L_0,L_1)$-Smoothness: Normalization and Momentum
- EvoBrain: Dynamic Multi-Channel EEG Graph Modeling for Time-Evolving Brain Networks
- Evolutionary Prediction Games
- Fading to Grow: Growing Preference Ratios via Preference Fading Discrete Diffusion for Recommendation
- FairNet: Dynamic Fairness Correction without Performance Loss via Contrastive Conditional LoRA
- FedSVD: Adaptive Orthogonalization for Private Federated Learning with LoRA
- FreeInv: Free Lunch for Improving DDIM Inversion
- From Black-box to Causal-box: Towards Building More Interpretable Models
- From Condensation to Rank Collapse: A Two-Stage Analysis of Transformer Training Dynamics
- From Condensation to Rank Collapse: A Two-Stage Analysis of Transformer Training Dynamics
- Fréchet Geodesic Boosting
- GPO: Learning from Critical Steps to Improve LLM Reasoning
- Geometric Imbalance in Semi-Supervised Node Classification
- Graph-Theoretic Insights into Bayesian Personalized Ranking for Recommendation
- HeavyWater and SimplexWater: Distortion-free LLM Watermarks for Low-Entropy Distributions
- High-order Interactions Modeling for Interpretable Multi-Agent Q-Learning
- Higher-Order Learning with Graph Neural Networks via Hypergraph Encodings
- How Does Topology Bias Distort Message Passing in Graph Recommender? A Dirichlet Energy Perspective
- How Many Domains Suffice for Domain Generalization? A Tight Characterization via the Domain Shattering Dimension
- How to Scale Second-Order Optimization
- Impact of Dataset Properties on Membership Inference Vulnerability of Deep Transfer Learning
- Improved Balanced Classification with Theoretically Grounded Loss Functions
- Infrequent Exploration in Linear Bandits
- Iterative Foundation Model Fine-Tuning on Multiple Rewards
- Joint Hierarchical Representation Learning of Samples and Features via Informed Tree-Wasserstein Distance
- KSP: Kolmogorov-Smirnov metric-based Post-Hoc Calibration for Survival Analysis
- LLM at Network Edge: A Layer-wise Efficient Federated Fine-tuning Approach
- LOPT: Learning Optimal Pigovian Tax in Sequential Social Dilemmas
- Learn2Mix: Training Neural Networks Using Adaptive Data Integration
- Learning Counterfactual Outcomes Under Rank Preservation
- Learning Reconfigurable Representations for Multimodal Federated Learning with Missing Data
- Learning Theory for Kernel Bilevel Optimization
- Learning from positive and unlabeled examples -Finite size sample bounds
- Leaving No OOD Instance Behind: Instance-Level OOD Fine-Tuning for Anomaly Segmentation
- Lifelong Test-Time Adaptation via Online Learning in Tracked Low-Dimensional Subspace
- MESS+: Dynamically Learned Inference-Time LLM Routing in Model Zoos with Service Level Guarantees
- Mamba Only Glances Once (MOGO): A Lightweight Framework for Efficient Video Action Detection
- Mixture of Scope Experts at Test: Generalizing Deeper Graph Neural Networks with Shallow Variants
- MoESD: Unveil Speculative Decoding's Potential for Accelerating Sparse MoE
- Multi-Objective Reinforcement Learning with Max-Min Criterion: A Game-Theoretic Approach
- MultiNet: Adaptive Multi-Viewed Subgraph Convolutional Networks for Graph Classification
- Multilevel neural simulation-based inference
- Multiplayer Federated Learning: Reaching Equilibrium with Less Communication
- Near-Optimal Sample Complexity for Online Constrained MDPs
- OCN: Effectively Utilizing Higher-Order Common Neighbors for Better Link Prediction
- OmniFC: Rethinking Federated Clustering via Lossless and Secure Distance Reconstruction
- On the $O(\frac{\sqrt{d}}{K^{1/4}})$ Convergence Rate of AdamW Measured by $\ell_1$ Norm
- On the Effect of Negative Gradient in Group Relative Deep Reinforcement Optimization
- On the Mechanisms of Weak-to-Strong Generalization: A Theoretical Perspective
- On the Optimality of the Median-of-Means Estimator under Adversarial Contamination
- On the SAC-BL Algorithm for Anomaly Detection
- On the Value of Cross-Modal Misalignment in Multimodal Representation Learning
- One SPACE to Rule Them All: Jointly Mitigating Factuality and Faithfulness Hallucinations in LLMs
- Optimal Dynamic Regret by Transformers for Non-Stationary Reinforcement Learning
- Optimal Mistake Bounds for Transductive Online Learning
- Optimistic Online-to-Batch Conversions for Accelerated Convergence and Universality
- Overcoming Long Context Limitations of State Space Models via Context Dependent Sparse Attention
- Pattern-Guided Adaptive Prior for Structure Learning
- Perturb a Model, Not an Image: Towards Robust Privacy Protection via Anti-Personalized Diffusion Models
- Pessimistic Data Integration for Policy Evaluation
- Polyline Path Masked Attention for Vision Transformer
- Predictability Enables Parallelization of Nonlinear State Space Models
- Predictive Preference Learning from Human Interventions
- Private Zeroth-Order Optimization with Public Data
- Problem-Parameter-Free Decentralized Bilevel Optimization
- Projective Equivariant Networks via Second-order Fundamental Differential Invariants
- Pruning Spurious Subgraphs for Graph Out-of-Distribution Generalization
- QuanDA: Quantile-Based Discriminant Analysis for High-Dimensional Imbalanced Classification
- RePO: Understanding Preference Learning Through ReLU-Based Optimization
- Reasoning Planning for Language Models
- Rescaled Influence Functions: Accurate Data Attribution in High Dimension
- Revisiting Glorot Initialization for Long-Range Linear Recurrences
- Revisiting Logit Distributions for Reliable Out-of-Distribution Detection
- Robust Federated Finetuning of LLMs via Alternating Optimization of LoRA
- Robust Graph Condensation via Classification Complexity Mitigation
- S'MoRE: Structural Mixture of Residual Experts for Parameter-Efficient LLM Fine-tuning
- STAIR: Addressing Stage Misalignment through Temporal-Aligned Preference Reinforcement Learning
- Scalable Fingerprinting of Large Language Models
- Sharp Analysis for KL-Regularized Contextual Bandits and RLHF
- SharpZO: Hybrid Sharpness-Aware Vision Language Model Prompt Tuning via Forward-Only Passes
- Sign-In to the Lottery: Reparameterizing Sparse Training
- Sparse Polyak: an adaptive step size rule for high-dimensional M-estimation
- Spectral Conditioning of Attention Improves Transformer Performance
- Spectral Learning for Infinite-Horizon Average-Reward POMDPs
- Stability and Sharper Risk Bounds with Convergence Rate $\tilde{O}(1/n^2)$
- Streaming Federated Learning with Markovian Data
- Synergy over Discrepancy: A Partition-Based Approach to Multi-Domain LLM Fine-Tuning
- T-REGS: Minimum Spanning Tree Regularization for Self-Supervised Learning
- Taming Hyperparameter Sensitivity in Data Attribution: Practical Selection Without Costly Retraining
- The Curse of Depth in Large Language Models
- The Graphon Limit Hypothesis: Understanding Neural Network Pruning via Infinite Width Analysis
- The emergence of sparse attention: impact of data distribution and benefits of repetition
- The emergence of sparse attention: impact of data distribution and benefits of repetition
- Theoretical Benefit and Limitation of Diffusion Language Model
- Through the River: Understanding the Benefit of Schedule-Free Methods for Language Model Training
- Thumb on the Scale: Optimal Loss Weighting in Last Layer Retraining
- Towards Self-Refinement of Vision-Language Models with Triangular Consistency
- Training-Free Bayesianization for Low-Rank Adapters of Large Language Models
- Transformers are almost optimal metalearners for linear classification
- Understanding Fairness and Prediction Error through Subspace Decomposition and Influence Analysis
- Understanding the Evolution of the Neural Tangent Kernel at the Edge of Stability
- Unveiling Extraneous Sampling Bias with Data Missing-Not-At-Random
- WHAT MAKES MATH PROBLEMS HARD FOR REINFORCEMENT LEARNING: A CASE STUDY
- Wasserstein Transfer Learning
- What Matters in Data for DPO?
- What Really is a Member? Discrediting Membership Inference via Poisoning
- Why Diffusion Models Don’t Memorize: The Role of Implicit Dynamical Regularization in Training
- Why Diffusion Models Don’t Memorize: The Role of Implicit Dynamical Regularization in Training
- Zero-Shot Blind-Spot Image Denoising via Cross-Scale Non-Local Pixel Refilling