Selected work

Representative Papers

Prediction-Oriented Subsampling from Data Streams

Aug 05, 2025

To address the challenge of balancing sampling efficiency and information preservation in offline learning under data stream settings, this paper proposes a prediction-oriented information-theoretic subsampling framework. Unlike conventional approaches that maximize input data entropy, our method guides sampling decisions by minimizing posterior uncertainty of downstream prediction tasks. It incorporates a lightweight model-aware mechanism to ensure sampling stability and computational tractability. Extensive experiments on time-series forecasting and anomaly detection demonstrate that the proposed method significantly outperforms existing information-theoretic baselines: it achieves an average 12.7% reduction in prediction error at equivalent sampling rates, while maintaining scalable computational overhead. The core contribution lies in the first explicit formulation of predictive uncertainty as a principled subsampling criterion—unifying theoretical interpretability with practical performance.

0 citationsRead paper

Distribution Transformers: Fast Approximate Bayesian Inference With On-The-Fly Prior Adaptation

Feb 04, 2025

Bayesian inference often relies on approximate methods due to intractable posterior computation; however, existing approaches are either computationally expensive or require retraining upon prior changes, hindering real-time sequential inference. This paper introduces the Distribution Transformer—a novel, learnable architecture capable of mapping arbitrary probability distributions. It is the first to uniformly represent both priors and posteriors using Gaussian Mixture Models (GMMs) and employs self-attention and cross-attention mechanisms for end-to-end distribution transformation, enabling online prior adaptation without retraining. Experiments demonstrate millisecond-scale inference latency—over 1,000× faster than conventional methods—while achieving log-likelihood performance competitive with or surpassing state-of-the-art baselines. The method is validated across diverse applications: sequential sensor fusion, quantum parameter estimation, and Gaussian process prediction with hyperpriors.

0 citationsRead paper

Recent publications

Latest Papers

Prediction-Oriented Subsampling from Data Streams

Aug 05, 2025

To address the challenge of balancing sampling efficiency and information preservation in offline learning under data stream settings, this paper proposes a prediction-oriented information-theoretic subsampling framework. Unlike conventional approaches that maximize input data entropy, our method guides sampling decisions by minimizing posterior uncertainty of downstream prediction tasks. It incorporates a lightweight model-aware mechanism to ensure sampling stability and computational tractability. Extensive experiments on time-series forecasting and anomaly detection demonstrate that the proposed method significantly outperforms existing information-theoretic baselines: it achieves an average 12.7% reduction in prediction error at equivalent sampling rates, while maintaining scalable computational overhead. The core contribution lies in the first explicit formulation of predictive uncertainty as a principled subsampling criterion—unifying theoretical interpretability with practical performance.

0 citationsRead paper

Distribution Transformers: Fast Approximate Bayesian Inference With On-The-Fly Prior Adaptation

Feb 04, 2025

Bayesian inference often relies on approximate methods due to intractable posterior computation; however, existing approaches are either computationally expensive or require retraining upon prior changes, hindering real-time sequential inference. This paper introduces the Distribution Transformer—a novel, learnable architecture capable of mapping arbitrary probability distributions. It is the first to uniformly represent both priors and posteriors using Gaussian Mixture Models (GMMs) and employs self-attention and cross-attention mechanisms for end-to-end distribution transformation, enabling online prior adaptation without retraining. Experiments demonstrate millisecond-scale inference latency—over 1,000× faster than conventional methods—while achieving log-likelihood performance competitive with or surpassing state-of-the-art baselines. The method is validated across diverse applications: sequential sensor fusion, quantum parameter estimation, and Gaussian process prediction with hyperpriors.

0 citationsRead paper