arXiv subjects

Yifeng Guo

Publications and source records attributed to Yifeng Guo.

9 recordsLinked to original sources

arXiv2026

Beyond Importance: Interchange-Sobol Sensitivity Reveals Task-Specific Content Channels in Transformer Components

Mechanistic interpretability methods summarize a transformer component by a single importance score, conflating two distinct roles: a component may matter because it transports task-relevant content, or because the forward computation degrades when its contribution is removed. We introduce \emph{Interchange-Group Sobol Decomposition} (IGSD), a paired-intervention framework that compares matched activation replacement with zero ablation on the same component, estimates two Sobol-style variance indices, and uses their signed difference to separate the two roles, with intervention validity monitored by a symmetric off-manifold diagnostic $\widehat{\mathrm{ST}}>1$. In factual recall, IGSD identifies an early-layer content channel in both GPT-2 small and Qwen2.5-1.5B that standard importance methods underestimate. A controlled subject and relation donor design shows that the early channel transports relation-frame content while late attention transports subject-retrieval content, refining at head granularity to the known $\mathrm{Attn}_{L9H8}$ head. Late-layer clamping confirms that the early signal is expressed through downstream transformations rather than residual pass-through. These results show that replacement and deletion are not interchangeable controls and their divergence provides a practical statistical diagnostic for content transport in transformer components.

stat.ML↗

arXiv2025

AI-Driven Prediction of Cancer Pain Episodes: A Hybrid Decision Support Approach

Lung cancer patients frequently experience breakthrough pain episodes, with up to 91% requiring timely intervention. To enable proactive pain management, we propose a hybrid machine learning and large language model pipeline that predicts pain episodes within 48 and 72 hours of hospitalization using both structured and unstructured electronic health record data. A retrospective cohort of 266 inpatients was analyzed, with features including demographics, tumor stage, vital signs, and WHO-tiered analgesic use. The machine learning module captured temporal medication trends, while the large language model interpreted ambiguous dosing records and free-text clinical notes. Integrating these modalities improved sensitivity and interpretability. Our framework achieved an accuracy of 0.876 (48h) and 0.917 (72h), with improvements in sensitivity of 10.6% and 10.7%, respectively, attributable to large language model augmentation. This hybrid approach offers a clinically interpretable and scalable tool for early pain episode forecasting, with potential to enhance treatment precision and optimize resource allocation in oncology care.

cs.AI↗

arXiv2024

ReverseNER: A Self-Generated Example-Driven Framework for Zero-Shot Named Entity Recognition with Large Language Models

This paper presents ReverseNER, a method aimed at overcoming the limitation of large language models (LLMs) in zero-shot named entity recognition (NER) tasks, arising from their reliance on pre-provided demonstrations. ReverseNER tackles this challenge by constructing a reliable example library composed of dozens of entity-labeled sentences, generated through the reverse process of NER. Specifically, while conventional NER methods label entities in a sentence, ReverseNER features reversing the process by using an LLM to generate entities from their definitions and subsequently expand them into full sentences. During the entity expansion process, the LLM is guided to generate sentences by replicating the structures of a set of specific \textsl{feature sentences}, extracted from the task sentences by clustering. This expansion process produces dozens of entity-labeled task-relevant sentences. After constructing the example library, the method selects several semantically similar entity-labeled examples for each task sentence as references to facilitate the LLM's entity recognition. We also propose an entity-level self-consistency scoring mechanism to improve NER performance with LLMs. Experiments show that ReverseNER significantly outperforms other zero-shot NER methods with LLMs, marking a notable improvement in NER for domains without labeled data, while declining computational resource consumption.

cs.CL↗

arXiv2023

HAR-Ito models and high-dimensional HAR modeling for high-frequency data

It is an important task to model realized volatilities for high-frequency data in finance and economics and, as arguably the most popular model, the heterogeneous autoregressive (HAR) model has dominated the applications in this area. However, this model suffers from three drawbacks: (i.) its heterogeneous volatility components are linear combinations of daily realized volatilities with fixed weights, which limit its flexibility for different types of assets, (ii.) it is still unknown what is the high-frequency probabilistic structure for this model, as well as many other HAR-type models in the literature, and (iii.) there is no high-dimensional inference tool for HAR modeling although it is common to encounter many assets in real applications. To overcome these drawbacks, this paper proposes a multilinear low-rank HAR model by using tensor techniques, where a data-driven method is adopted to automatically select the heterogeneous components. In addition, HAR-Itô models are introduced to interpret the corresponding high-frequency dynamics, as well as those of other HAR-type models. Moreover, non-asymptotic properties of the high-dimensional HAR modeling are established, and a projected gradient descent algorithm with theoretical justifications is suggested to search for estimates. Theoretical and computational properties of the proposed method are verified by simulation studies, and the necessity of using the data-driven method for heterogeneous components is illustrated in real data analysis.

stat.ME↗

arXiv2023

Hierarchical Perception Adversarial Learning Framework for Compressed Sensing MRI

The long acquisition time has limited the accessibility of magnetic resonance imaging (MRI) because it leads to patient discomfort and motion artifacts. Although several MRI techniques have been proposed to reduce the acquisition time, compressed sensing in magnetic resonance imaging (CS-MRI) enables fast acquisition without compromising SNR and resolution. However, existing CS-MRI methods suffer from the challenge of aliasing artifacts. This challenge results in the noise-like textures and missing the fine details, thus leading to unsatisfactory reconstruction performance. To tackle this challenge, we propose a hierarchical perception adversarial learning framework (HP-ALF). HP-ALF can perceive the image information in the hierarchical mechanism: image-level perception and patch-level perception. The former can reduce the visual perception difference in the entire image, and thus achieve aliasing artifact removal. The latter can reduce this difference in the regions of the image, and thus recover fine details. Specifically, HP-ALF achieves the hierarchical mechanism by utilizing multilevel perspective discrimination. This discrimination can provide the information from two perspectives (overall and regional) for adversarial learning. It also utilizes a global and local coherent discriminator to provide structure information to the generator during training. In addition, HP-ALF contains a context-aware learning block to effectively exploit the slice information between individual images for better reconstruction performance. The experiments validated on three datasets demonstrate the effectiveness of HP-ALF and its superiority to the comparative methods.

eess.IV↗

arXiv2022

High Dimensional Portfolio Selection with Cardinality Constraints

The expanding number of assets offers more opportunities for investors but poses new challenges for modern portfolio management (PM). As a central plank of PM, portfolio selection by expected utility maximization (EUM) faces uncontrollable estimation and optimization errors in ultrahigh-dimensional scenarios. Past strategies for high-dimensional PM mainly concern only large-cap companies and select many stocks, making PM impractical. We propose a sample-average approximation-based portfolio strategy to tackle the difficulties above with cardinality constraints. Our strategy bypasses the estimation of mean and covariance, the Chinese walls in high-dimensional scenarios. Empirical results on S&P 500 and Russell 2000 show that an appropriate number of carefully chosen assets leads to better out-of-sample mean-variance efficiency. On Russell 2000, our best portfolio profits as much as the equally-weighted portfolio but reduces the maximum drawdown and the average number of assets by 10% and 90%, respectively. The flexibility and the stability of incorporating factor signals for augmenting out-of-sample performances are also demonstrated. Our strategy balances the trade-off among the return, the risk, and the number of assets with cardinality constraints. Therefore, we provide a theoretically sound and computationally efficient strategy to make PM practical in the growing global financial market.

stat.AP↗

arXiv2020

Explainable Recommendation Systems by Generalized Additive Models with Manifest and Latent Interactions

In recent years, the field of recommendation systems has attracted increasing attention to developing predictive models that provide explanations of why an item is recommended to a user. The explanations can be either obtained by post-hoc diagnostics after fitting a relatively complex model or embedded into an intrinsically interpretable model. In this paper, we propose the explainable recommendation systems based on a generalized additive model with manifest and latent interactions (GAMMLI). This model architecture is intrinsically interpretable, as it additively consists of the user and item main effects, the manifest user-item interactions based on observed features, and the latent interaction effects from residuals. Unlike conventional collaborative filtering methods, the group effect of users and items are considered in GAMMLI. It is beneficial for enhancing the model interpretability, and can also facilitate the cold-start recommendation problem. A new Python package GAMMLI is developed for efficient model training and visualized interpretation of the results. By numerical experiments based on simulation data and real-world cases, the proposed method is shown to have advantages in both predictive performance and explainable recommendation.

cs.LG↗

arXiv2020

Deep Attentive Wasserstein Generative Adversarial Networks for MRI Reconstruction with Recurrent Context-Awareness

The performance of traditional compressive sensing-based MRI (CS-MRI) reconstruction is affected by its slow iterative procedure and noise-induced artefacts. Although many deep learning-based CS-MRI methods have been proposed to mitigate the problems of traditional methods, they have not been able to achieve more robust results at higher acceleration factors. Most of the deep learning-based CS-MRI methods still can not fully mine the information from the k-space, which leads to unsatisfactory results in the MRI reconstruction. In this study, we propose a new deep learning-based CS-MRI reconstruction method to fully utilise the relationship among sequential MRI slices by coupling Wasserstein Generative Adversarial Networks (WGAN) with Recurrent Neural Networks. Further development of an attentive unit enables our model to reconstruct more accurate anatomical structures for the MRI data. By experimenting on different MRI datasets, we have demonstrated that our method can not only achieve better results compared to the state-of-the-arts but can also effectively reduce residual noise generated during the reconstruction process.

eess.IV↗

arXiv2018

Robust Log-Optimal Strategy with Reinforcement Learning

We proposed a new Portfolio Management method termed as Robust Log-Optimal Strategy (RLOS), which ameliorates the General Log-Optimal Strategy (GLOS) by approximating the traditional objective function with quadratic Taylor expansion. It avoids GLOS's complex CDF estimation process,hence resists the "Butterfly Effect" caused by estimation error. Besides,RLOS retains GLOS's profitability and the optimization problem involved in RLOS is computationally far more practical compared to GLOS. Further, we combine RLOS with Reinforcement Learning (RL) and propose the so-called Robust Log-Optimal Strategy with Reinforcement Learning (RLOSRL), where the RL agent receives the analyzed results from RLOS and observes the trading environment to make comprehensive investment decisions. The RLOSRL's performance is compared to some traditional strategies on several back tests, where we randomly choose a selection of constituent stocks of the CSI300 index as assets under management and the test results validate its profitability and stability.

q-fin.PM↗