Machine Learning

Fine-tuning pre-trained models often degrades performance on unseen classes. This work reveals that the problem stems from logit scale discrepancies, not feature loss, and shows that post-processing c…

A new fine-grained dynamic framework jointly optimizes bias and variance for accurate predictions from missing-not-at-random data, surpassing existing methods.

First near-optimal reinforcement learning algorithm achieving best policy identification in average-reward MDPs without prior knowledge of complexity.

FilterNet: A novel deep learning architecture using learnable frequency filters for superior time series forecasting accuracy and efficiency.

FIDE, a novel conditional diffusion model, accurately generates time series by inflating high-frequency components, preserving extreme value distributions.

FIARSE dynamically optimizes submodels in federated learning based on parameter importance, improving efficiency and global model accuracy.

Few-shot task learning through inverse generative modeling (FTL-IGM) enables AI agents to quickly master new tasks from minimal data by leveraging invertible generative models.

Few-shot diffusion models efficiently generate customized images; this paper provides the first theoretical explanation, proving improved approximation and optimization bounds, escaping the curse of d…

Ferrari, a novel federated feature unlearning framework, minimizes feature sensitivity via Lipschitz continuity, enabling effective and privacy-preserving data removal without full client participatio…

FedSSP tackles personalized federated graph learning challenges by sharing generic spectral knowledge and incorporating personalized preferences, achieving superior performance in cross-domain scenari…

FEDNE: a novel approach enabling collaborative dimensionality reduction of distributed data in federated learning without data sharing, achieved via surrogate loss functions and data augmentation.

FedLPA: One-shot federated learning with layer-wise posterior aggregation improves model accuracy in non-IID data by efficiently aggregating layer-wise posteriors of local models using a novel approac…

FedGTST significantly improves federated transfer learning by tuning cross-client statistics, achieving superior global transferability with minimal communication overhead.

FedGMKD: A novel federated learning framework uses knowledge distillation and discrepancy-aware aggregation for efficient, privacy-preserving personalized learning in heterogeneous data settings.

FedGMark: the first certified robust watermarking method for protecting Federated Graph Learning models against theft and unauthorized copying.

Federated Transformer (FeT) revolutionizes multi-party fuzzy vertical federated learning by encoding fuzzy identifiers and using a transformer architecture, achieving up to 46% accuracy improvement an…

This paper presents novel algorithms achieving speed-ups in differentially private federated online prediction from experts, addressing both stochastic and oblivious adversaries, with rigorous theoret…

This paper introduces federated natural policy gradient and actor-critic methods achieving near dimension-free global convergence for decentralized multi-task reinforcement learning, a significant bre…

FedMRL: a novel federated learning approach achieves high accuracy with low communication cost by enabling clients with heterogeneous models to collaboratively train using shared auxiliary models and …

Amplified SCAFFOLD: A new algorithm for federated learning significantly reduces communication rounds under periodic client participation and heterogeneous data, achieving linear speedup and resilienc…