arXiv subjects

Kewen Liao

Publications and source records attributed to Kewen Liao.

16 recordsLinked to original sources

arXiv2026

CoAuthorAI: A Human in the Loop System For Scientific Book Writing

Large language models (LLMs) are increasingly used in scientific writing but struggle with book-length tasks, often producing inconsistent structure and unreliable citations. We introduce CoAuthorAI, a human-in-the-loop writing system that combines retrieval-augmented generation, expert-designed hierarchical outlines, and automatic reference linking. The system allows experts to iteratively refine text at the sentence level, ensuring coherence and accuracy. In evaluations of 500 multi-domain literature review chapters, CoAuthorAI achieved a maximum soft-heading recall of 98%; in a human evaluation of 100 articles, the generated content reached a satisfaction rate of 82%. The book AI for Rock Dynamics generated with CoAuthorAI and Kexin Technology's LUFFA AI model has been published with Springer Nature. These results show that systematic human-AI collaboration can extend LLMs' capabilities from articles to full-length books, enabling faster and more reliable scientific publishing.

cs.CL↗

arXiv2026

Approximation Algorithm for Constrained $k$-Center Clustering: A Local Search Approach

Clustering is a long-standing research problem and a fundamental tool in AI and data analysis. The traditional k-center problem, a fundamental theoretical challenge in clustering, has a best possible approximation ratio of 2, and any improvement to a ratio of 2 - {\epsilon} would imply P = NP. In this work, we study the constrained k-center clustering problem, where instance-level cannot-link (CL) and must-link (ML) constraints are incorporated as background knowledge. Although general CL constraints significantly increase the hardness of approximation, previous work has shown that disjoint CL sets permit constant-factor approximations. However, whether local search can achieve such a guarantee in this setting remains an open question. To this end, we propose a novel local search framework based on a transformation to a dominating matching set problem, achieving the best possible approximation ratio of 2. The experimental results on both real-world and synthetic datasets demonstrate that our algorithm outperforms baselines in solution quality.

cs.LG↗

arXiv2025

TED++: Submanifold-Aware Backdoor Detection via Layerwise Tubular-Neighbourhood Screening

As deep neural networks power increasingly critical applications, stealthy backdoor attacks, where poisoned training inputs trigger malicious model behaviour while appearing benign, pose a severe security risk. Many existing defences are vulnerable when attackers exploit subtle distance-based anomalies or when clean examples are scarce. To meet this challenge, we introduce TED++, a submanifold-aware framework that effectively detects subtle backdoors that evade existing defences. TED++ begins by constructing a tubular neighbourhood around each class's hidden-feature manifold, estimating its local ``thickness'' from a handful of clean activations. It then applies Locally Adaptive Ranking (LAR) to detect any activation that drifts outside the admissible tube. By aggregating these LAR-adjusted ranks across all layers, TED++ captures how faithfully an input remains on the evolving class submanifolds. Based on such characteristic ``tube-constrained'' behaviour, TED++ flags inputs whose LAR-based ranking sequences deviate significantly. Extensive experiments are conducted on benchmark datasets and tasks, demonstrating that TED++ achieves state-of-the-art detection performance under both adaptive-attack and limited-data scenarios. Remarkably, even with only five held-out examples per class, TED++ still delivers near-perfect detection, achieving gains of up to 14\% in AUROC over the next-best method. The code is publicly available at https://github.com/namle-w/TEDpp.

cs.LG↗

arXiv2025

Looking in the mirror: A faithful counterfactual explanation method for interpreting deep image classification models

Counterfactual explanations (CFE) for deep image classifiers aim to reveal how minimal input changes lead to different model decisions, providing critical insights for model interpretation and improvement. However, existing CFE methods often rely on additional image encoders and generative models to create plausible images, neglecting the classifier's own feature space and decision boundaries. As such, they do not explain the intrinsic feature space and decision boundaries learned by the classifier. To address this limitation, we propose Mirror-CFE, a novel method that generates faithful counterfactual explanations by operating directly in the classifier's feature space, treating decision boundaries as mirrors that ``reflect'' feature representations in the mirror. Mirror-CFE learns a mapping function from feature space to image space while preserving distance relationships, enabling smooth transitions between source images and their counterfactuals. Through extensive experiments on four image datasets, we demonstrate that Mirror-CFE achieves superior performance in validity while maintaining input resemblance compared to state-of-the-art explanation methods. Finally, mirror-CFE provides interpretable visualization of the classifier's decision process by generating step-wise transitions that reveal how features evolve as classification confidence changes.

cs.CV↗

arXiv2025

DeepFeatIoT: Unifying Deep Learned, Randomized, and LLM Features for Enhanced IoT Time Series Sensor Data Classification in Smart Industries

Internet of Things (IoT) sensors are ubiquitous technologies deployed across smart cities, industrial sites, and healthcare systems. They continuously generate time series data that enable advanced analytics and automation in industries. However, challenges such as the loss or ambiguity of sensor metadata, heterogeneity in data sources, varying sampling frequencies, inconsistent units of measurement, and irregular timestamps make raw IoT time series data difficult to interpret, undermining the effectiveness of smart systems. To address these challenges, we propose a novel deep learning model, DeepFeatIoT, which integrates learned local and global features with non-learned randomized convolutional kernel-based features and features from large language models (LLMs). This straightforward yet unique fusion of diverse learned and non-learned features significantly enhances IoT time series sensor data classification, even in scenarios with limited labeled data. Our model's effectiveness is demonstrated through its consistent and generalized performance across multiple real-world IoT sensor datasets from diverse critical application domains, outperforming state-of-the-art benchmark models. These results highlight DeepFeatIoT's potential to drive significant advancements in IoT analytics and support the development of next-generation smart systems.

cs.LG↗

arXiv2024

AdaCBM: An Adaptive Concept Bottleneck Model for Explainable and Accurate Diagnosis

The integration of vision-language models such as CLIP and Concept Bottleneck Models (CBMs) offers a promising approach to explaining deep neural network (DNN) decisions using concepts understandable by humans, addressing the black-box concern of DNNs. While CLIP provides both explainability and zero-shot classification capability, its pre-training on generic image and text data may limit its classification accuracy and applicability to medical image diagnostic tasks, creating a transfer learning problem. To maintain explainability and address transfer learning needs, CBM methods commonly design post-processing modules after the bottleneck module. However, this way has been ineffective. This paper takes an unconventional approach by re-examining the CBM framework through the lens of its geometrical representation as a simple linear classification system. The analysis uncovers that post-CBM fine-tuning modules merely rescale and shift the classification outcome of the system, failing to fully leverage the system's learning potential. We introduce an adaptive module strategically positioned between CLIP and CBM to bridge the gap between source and downstream domains. This simple yet effective approach enhances classification performance while preserving the explainability afforded by the framework. Our work offers a comprehensive solution that encompasses the entire process, from concept discovery to model training, providing a holistic recipe for leveraging the strengths of GPT, CLIP, and CBM.

cs.CV↗

arXiv2024

CAPE: CAM as a Probabilistic Ensemble for Enhanced DNN Interpretation

Deep Neural Networks (DNNs) are widely used for visual classification tasks, but their complex computation process and black-box nature hinder decision transparency and interpretability. Class activation maps (CAMs) and recent variants provide ways to visually explain the DNN decision-making process by displaying 'attention' heatmaps of the DNNs. Nevertheless, the CAM explanation only offers relative attention information, that is, on an attention heatmap, we can interpret which image region is more or less important than the others. However, these regions cannot be meaningfully compared across classes, and the contribution of each region to the model's class prediction is not revealed. To address these challenges that ultimately lead to better DNN Interpretation, in this paper, we propose CAPE, a novel reformulation of CAM that provides a unified and probabilistically meaningful assessment of the contributions of image regions. We quantitatively and qualitatively compare CAPE with state-of-the-art CAM methods on CUB and ImageNet benchmark datasets to demonstrate enhanced interpretability. We also test on a cytology imaging dataset depicting a challenging Chronic Myelomonocytic Leukemia (CMML) diagnosis problem. Code is available at: https://github.com/AIML-MED/CAPE.

cs.CV↗

arXiv2024

DeepHeteroIoT: Deep Local and Global Learning over Heterogeneous IoT Sensor Data

Internet of Things (IoT) sensor data or readings evince variations in timestamp range, sampling frequency, geographical location, unit of measurement, etc. Such presented sequence data heterogeneity makes it difficult for traditional time series classification algorithms to perform well. Therefore, addressing the heterogeneity challenge demands learning not only the sub-patterns (local features) but also the overall pattern (global feature). To address the challenge of classifying heterogeneous IoT sensor data (e.g., categorizing sensor data types like temperature and humidity), we propose a novel deep learning model that incorporates both Convolutional Neural Network and Bi-directional Gated Recurrent Unit to learn local and global features respectively, in an end-to-end manner. Through rigorous experimentation on heterogeneous IoT sensor datasets, we validate the effectiveness of our proposed model, which outperforms recent state-of-the-art classification methods as well as several machine learning and deep learning baselines. In particular, the model achieves an average absolute improvement of 3.37% in Accuracy and 2.85% in F1-Score across datasets

cs.LG↗

arXiv2024

Near-Optimal Algorithms for Constrained k-Center Clustering with Instance-level Background Knowledge

Center-based clustering has attracted significant research interest from both theory and practice. In many practical applications, input data often contain background knowledge that can be used to improve clustering results. In this work, we build on widely adopted $k$-center clustering and model its input background knowledge as must-link (ML) and cannot-link (CL) constraint sets. However, most clustering problems including $k$-center are inherently $\mathcal{NP}$-hard, while the more complex constrained variants are known to suffer severer approximation and computation barriers that significantly limit their applicability. By employing a suite of techniques including reverse dominating sets, linear programming (LP) integral polyhedron, and LP duality, we arrive at the first efficient approximation algorithm for constrained $k$-center with the best possible ratio of 2. We also construct competitive baseline algorithms and empirically evaluate our approximation algorithm against them on a variety of real datasets. The results validate our theoretical findings and demonstrate the great advantages of our algorithm in terms of clustering cost, clustering quality, and running time.

cs.LG↗

arXiv2022

CNN Attention Guidance for Improved Orthopedics Radiographic Fracture Classification

Convolutional neural networks (CNNs) have gained significant popularity in orthopedic imaging in recent years due to their ability to solve fracture classification problems. A common criticism of CNNs is their opaque learning and reasoning process, making it difficult to trust machine diagnosis and the subsequent adoption of such algorithms in clinical setting. This is especially true when the CNN is trained with limited amount of medical data, which is a common issue as curating sufficiently large amount of annotated medical imaging data is a long and costly process. While interest has been devoted to explaining CNN learnt knowledge by visualizing network attention, the utilization of the visualized attention to improve network learning has been rarely investigated. This paper explores the effectiveness of regularizing CNN network with human-provided attention guidance on where in the image the network should look for answering clues. On two orthopedics radiographic fracture classification datasets, through extensive experiments we demonstrate that explicit human-guided attention indeed can direct correct network attention and consequently significantly improve classification performance. The development code for the proposed attention guidance is publicly available on GitHub.

cs.CV↗

arXiv2021

CorrDetector: A Framework for Structural Corrosion Detection from Drone Images using Ensemble Deep Learning

In this paper, we propose a new technique that applies automated image analysis in the area of structural corrosion monitoring and demonstrate improved efficacy compared to existing approaches. Structural corrosion monitoring is the initial step of the risk-based maintenance philosophy and depends on an engineer's assessment regarding the risk of building failure balanced against the fiscal cost of maintenance. This introduces the opportunity for human error which is further complicated when restricted to assessment using drone captured images for those areas not reachable by humans due to many background noises. The importance of this problem has promoted an active research community aiming to support the engineer through the use of artificial intelligence (AI) image analysis for corrosion detection. In this paper, we advance this area of research with the development of a framework, CorrDetector. CorrDetector uses a novel ensemble deep learning approach underpinned by convolutional neural networks (CNNs) for structural identification and corrosion feature extraction. We provide an empirical evaluation using real-world images of a complicated structure (e.g. telecommunication tower) captured by drones, a typical scenario for engineers. Our study demonstrates that the ensemble approach of \model significantly outperforms the state-of-the-art in terms of classification accuracy.

cs.CV↗

arXiv2015

Efficient Approximation Algorithms for Computing \emph{k} Disjoint Restricted Shortest Paths

Network applications, such as multimedia streaming and video conferencing, impose growing requirements over Quality of Service (QoS), including bandwidth, delay, jitter, etc. Meanwhile, networks are expected to be load-balanced, energy-efficient, and resilient to some degree of failures. It is observed that the above requirements could be better met with multiple disjoint QoS paths than a single one. Let $G=(V,\, E)$ be a digraph with nonnegative integral cost and delay on every edge, $s,\, t\in V$ be two specified vertices, and $D\in\mathbb{Z}_{0}^{+}$ be a delay bound (or some other constraint), the \emph{$k$ Disjoint Restricted Shortest Path} ($k$\emph{RSP})\emph{ Problem} is computing $k$ disjoint paths between $s$ and $t$ with total cost minimized and total delay bounded by $D$. Few efficient algorithms have been developed because of the hardness of the problem. In this paper, we propose efficient algorithms with provable performance guarantees for the $k$RSP problem. We first present a pseudo-polynomial-time approximation algorithm with a bifactor approximation ratio of $(1,\,2)$, then improve the algorithm to polynomial time with a bifactor ratio of $(1+ε,\,2+ε)$ for any fixed $ε>0$, which is better than the current best approximation ratio $(O(1+γ),\, O(1+\frac{1}γ)\})$ for any fixed $γ>0$ \cite{orda2004efficient}. To the best of our knowledge, this is the first constant-factor algorithm that almost strictly obeys the constraint for the $k$RSP problem.

cs.DM↗

arXiv2013

A Parameterized Approximation Algorithm for The Shallow-Light Steiner Tree Problem

For a given graph $G=(V,\, E)$ with a terminal set $S$ and a selected root $r\in S$, a positive integer cost and a delay on every edge and a delay constraint $D\in Z^{+}$, the shallow-light Steiner tree (\emph{SLST}) problem is to compute a minimum cost tree spanning the terminals of $S$, in which the delay between root and every vertex is restrained by $D$. This problem is NP-hard and very hard to approximate. According to known inapproximability results, this problem admits no approximation with ratio better than factor $(1,\, O(\log^{2}n))$ unless $NP\subseteq DTIME(n^{\log\log n})$ \cite{khandekar2013some}, while it admits no approximation ratio better than $(1,\, O(\log|V|))$ for D=4 unless $NP\subseteq DTIME(n^{\log\log n})$ \cite{bar2001generalized}. Hence, the paper focus on parameterized algorithm for \emph{SLST}. We firstly present an exact algorithm for \emph{SLST} with time complexity $O(3^{|S|}|V|D+2^{|S|}|V|^{2}D^{2}+|V|^{3}D^{3})$, where $|S|$ and $|V|$ are the number of terminals and vertices respectively. This is a pseudo polynomial time parameterized algorithm with respect to the parameterization: "number of terminals". Later, we improve this algorithm such that it runs in polynomial time $O(\frac{|V|^{2}}ε3^{|S|}+\frac{|V|^{4}}ε2^{|S|}+\frac{|V|^{6}}ε)$, and computes a Steiner tree with delay bounded by $(1+ε)D$ and cost bounded by the cost of an optimum solution, where $ε>0$ is any small real number. To the best of our knowledge, this is the first parameterized approximation algorithm for the \emph{SLST} problem.

cs.DS↗

arXiv2013

Constrained Fault-Tolerant Resource Allocation

In the Constrained Fault-Tolerant Resource Allocation (FTRA) problem, we are given a set of sites containing facilities as resources, and a set of clients accessing these resources. Specifically, each site i is allowed to open at most R_i facilities with cost f_i for each opened facility. Each client j requires an allocation of r_j open facilities and connecting j to any facility at site i incurs a connection cost c_ij. The goal is to minimize the total cost of this resource allocation scenario. FTRA generalizes the Unconstrained Fault-Tolerant Resource Allocation (FTRA_{\infty}) [18] and the classical Fault-Tolerant Facility Location (FTFL) [13] problems: for every site i, FTRA_{\infty} does not have the constraint R_i, whereas FTFL sets R_i=1. These problems are said to be uniform if all r_j's are the same, and general otherwise. For the general metric FTRA, we first give an LP-rounding algorithm achieving the approximation ratio of 4. Then we show the problem reduces to FTFL, implying the ratio of 1.7245 from [3]. For the uniform FTRA, we provide a 1.52-approximation primal-dual algorithm in O(n^4) time, where n is the total number of sites and clients. We also consider the Constrained Fault-Tolerant k-Resource Allocation (k-FTRA) problem where additionally the total number of facilities can be opened across all sites is bounded by k. For the uniform k-FTRA, we give the first constant-factor approximation algorithm with a factor of 4. Note that the above results carry over to FTRA_{\infty} and k-FTRA_{\infty}.

cs.DS↗

arXiv2013

Improved Approximation Algorithms for Computing k Disjoint Paths Subject to Two Constraints

For a given graph $G$ with positive integral cost and delay on edges, distinct vertices $s$ and $t$, cost bound $C\in Z^{+}$ and delay bound $D\in Z^{+}$, the $k$ bi-constraint path ($k$BCP) problem is to compute $k$ disjoint $st$-paths subject to $C$ and $D$. This problem is known NP-hard, even when $k=1$ \cite{garey1979computers}. This paper first gives a simple approximation algorithm with factor-$(2,2)$, i.e. the algorithm computes a solution with delay and cost bounded by $2*D$ and $2*C$ respectively. Later, a novel improved approximation algorithm with ratio $(1+β,\,\max\{2,\,1+\ln\frac{1}β\})$ is developed by constructing interesting auxiliary graphs and employing the cycle cancellation method. As a consequence, we can obtain a factor-$(1.369,\,2)$ approximation algorithm by setting $1+\ln\frac{1}β=2$ and a factor-$(1.567,\,1.567)$ algorithm by setting $1+β=1+\ln\frac{1}β$. Besides, by setting $β=0$, an approximation algorithm with ratio $(1,\, O(\ln n))$, i.e. an algorithm with only a single factor ratio $O(\ln n)$ on cost, can be immediately obtained. To the best of our knowledge, this is the first non-trivial approximation algorithm for the $k$BCP problem that strictly obeys the delay constraint.

cs.DS↗

arXiv2011

Unconstrained and Constrained Fault-Tolerant Resource Allocation

First, we study the Unconstrained Fault-Tolerant Resource Allocation (UFTRA) problem (a.k.a. FTFA problem in \cite{shihongftfa}). In the problem, we are given a set of sites equipped with an unconstrained number of facilities as resources, and a set of clients with set $\mathcal{R}$ as corresponding connection requirements, where every facility belonging to the same site has an identical opening (operating) cost and every client-facility pair has a connection cost. The objective is to allocate facilities from sites to satisfy $\mathcal{R}$ at a minimum total cost. Next, we introduce the Constrained Fault-Tolerant Resource Allocation (CFTRA) problem. It differs from UFTRA in that the number of resources available at each site $i$ is limited by $R_{i}$. Both problems are practical extensions of the classical Fault-Tolerant Facility Location (FTFL) problem \cite{Jain00FTFL}. For instance, their solutions provide optimal resource allocation (w.r.t. enterprises) and leasing (w.r.t. clients) strategies for the contemporary cloud platforms. In this paper, we consider the metric version of the problems. For UFTRA with uniform $\mathcal{R}$, we present a star-greedy algorithm. The algorithm achieves the approximation ratio of 1.5186 after combining with the cost scaling and greedy augmentation techniques similar to \cite{Charikar051.7281.853,Mahdian021.52}, which significantly improves the result of \cite{shihongftfa} using a phase-greedy algorithm. We also study the capacitated extension of UFTRA and give a factor of 2.89. For CFTRA with uniform $\mathcal{R}$, we slightly modify the algorithm to achieve 1.5186-approximation. For a more general version of CFTRA, we show that it is reducible to FTFL using linear programming.

cs.DS↗