共找到 20 条结果
An 81-year-old man with a history of multiple myeloma presented with decompensated heart failure (HF). Although his baseline left ventricular ejection fraction (LVEF) was preserved, it significantly declined to 35% at 13 months after the initiation of ixazomib. Considering the chronological relationship between the timing of ixazomib treatment and symptom onset, prior normal echocardiography, and multimodal imaging findings, ixazomib-induced cancer therapy-related cardiac dysfunction (CTRCD) was strongly suspected. Despite the discontinuation of ixazomib and continued administration of guideline-directed medical therapy, his LVEF has not recovered to date. We herein report a rare case of ixazomib-induced irreversible CTRCD.
暂无摘要（点击查看详情）
This study aims to optimize the [Formula: see text]-nearest neighbors search (kNN search) by reducing the computational burden of the well-known Brute-force method while providing the same solution. While there exist rule-based approaches for reducing the computational burden of the kNN search, methods that use the stochastic patterns inherent to the data are lacking. Our method leverages data structures and probabilistic assumptions to enhance the scalability of the search. By focusing on the Training set where our neighbors reside, we define a sample space that limits the [Formula: see text]-nearest neighbors search to a smaller space. For each observation in the Query set (e.g., the set of observations for which a classification is desired), a fixed radius search is employed, with the radius stochastically linked to the desired number of neighbors. This approach allows us to find the [Formula: see text]-nearest neighbors using only a fraction of the entire Training set in contrast to the Brute-force method, which requires distances to be calculated between each observation in the Training set and each observation in the Query set. Through simulations and a theoretical computational complexity analysis, we demonstrate that our method outperforms the Brute-force approach, particularly when the Training and Query set sample sizes are large. In addition, a benchmarked comparison of our approach and the Brute-force method on an Alzheimer's disease data set further demonstrated this, showing a 27.57-fold improvement in total elapsed time. Overall, our stochastic approach significantly reduces the computational load of kNN search while maintaining accuracy, making it a viable alternative to traditional methods for large datasets.
We propose the integral-equation formalism of population dynamics (IEPDYN) to describe the population dynamics of distinct configurational states. According to classical reaction dynamics theory, the probability density associated with a given state obeys the Liouville equation, including influx from and efflux to neighboring states. By introducing a Markov approximation for the crossing of boundaries separating the states, tractable integral equations governing the state populations are derived. Once the time-dependent quantities appearing in these equations are evaluated, the population dynamics on long timescales can be obtained. Because these quantities depend only on a few states in the local neighborhood of a given state, they can be computed using a set of short-timescale molecular dynamics (MD) simulations. The IEPDYN method is formulated in continuous time and therefore does not rely on a coarse-grained timescale (lag time). Consequently, kinetic quantities obtained from IEPDYN are free from lag-time dependence, which has been discussed as a limitation in other approaches. We apply the IEPDYN method to the binding and unbinding kinetics of CH4/CH4, Na+/Cl-, and 18-crown-6-ether (crown ether)/K+ in water. For both kinetics, the time constants estimated from the IEPDYN method are comparable to those obtained from brute-force MD simulations. The required timescale of each MD trajectory in the IEPDYN method is approximately two orders of magnitude shorter than that in the brute-force MD approach in the crown ether/K+ system. This reduction in the trajectory timescale enables applications to complex binding and unbinding systems whose characteristic timescales are far beyond those directly accessible by brute-force MD simulations.
Limited experimental data remains a key challenge in applying machine learning to drug discovery, particularly for cancer-related targets. In this study, we present a data-efficient active meta-deep learning framework to predict mitogen-activated protein kinase 1 (MAPK1) inhibitors, which are promising candidates for cancer-related therapies. Our approach integrates active learning (AL) with a meta-model that combines four deep architectures: a convolutional neural network, an attention, a graph convolutional network, and a graph neural network-attention, trained on molecular descriptors and graph-based representations. These models generate four probability-based features that feed into an attention-based meta-learner, improving predictive performance by 5.12% in the area under the precision-recall curve (AUPRC) and 5.48% in the Matthews correlation coefficient (MCC) using only 10% of the training data. Among the AL sampling strategies evaluated, entropy sampling showed competitive performance in selecting informative molecules for model improvement. Overall, our framework achieves an AUPRC of 0.835 ± 0.017 and MCC of 0.817 ± 0.017, on par with a traditional training method despite using only 26.7% of the training data. Compared to a conventional random forest model trained on brute-force, a 100% full training set, our approach shows a 10.6% improvement in AUPRC and modest gains in MCC, confirming the effectiveness of the proposed framework. Under severe class imbalance, balanced accuracy steadily increased across AL iterations, reaching values greater than 0.85 at the final iteration for all uncertainty-driven strategies. Molecular docking confirmed successful prioritization of the top four predicted compounds. Evaluation on an external MAPK1 data set demonstrated generalizability, with our approach achieving an AUPRC of 0.818 and an MCC of 0.403, comparable to the independent test set. These results highlight the potential of combining intelligent data selection with deep learning architectures through the meta-model to accelerate predictive performance in data-scarce drug discovery. Scientific contribution: This study contributes a novel, data-efficient active meta-deep learning framework for predicting MAPK1 inhibitors, addressing the challenge of limited experimental data in a cancer-specific target. By integrating AL with a meta-model composed of four deep architectures, the approach significantly enhances the predictive performance using only a fraction of the training data. The framework achieves superior metrics compared to traditional training methods, highlighting its potential to accelerate drug discovery in data-scarce settings.
This paper presents NAVBOT25, a labelled dataset aimed at strengthening the overall security of autonomous robots against network-based cyber threats within the Robot Operating System (ROS) platform. NAVBOT25 offers a comprehensive, labelled dataset that captures both normal operational behaviour and a variety of attack scenarios relevant to ROS-based systems. This dataset was generated by deploying a TurtleBot3 running ROS Noetic in controlled laboratory setting, where real-world attack vectors were executed -including SSH brute-force attempts, reverse shells, port scans, and ROS-specific attacks such as unauthorized publishing actions and topic flooding. Network traffic was captured using tcpdump, and 83 flow-level features were extracted using CICFlowMeter, resulting in a series of CSV files. Designed to support the development of AI-assisted intrusion detection systems, NAVBOT25 addresses existing gaps in robotic cybersecurity research by providing a richer and more diverse dataset for evaluating threat detection in networked robotic systems.
Basecalling is a crucial step in DNA sequencing that converts raw nanopore signals into nucleotide sequences. This paper presents a serial-parallel reprogrammable DNA sequencing accelerator based on a 64-state Hidden Markov Model (HMM) implemented in a 130-nm CMOS process. The proposed method optimizes computational efficiency, hardware utilization, and power consumption using a coarse-grained serial-parallel processing approach. It achieves 94.3% accuracy, outperforming Nanocall (85.6%) and Meta-Align (91.2%), while being slightly superior to the Scalable Hardware Accelerator (93.1%). Furthermore, it consumes 200 mW, which is 6 times lower than brute-force HMM implementations and 3–5 times more power-efficient than deep learning-based basecallers like DeepNano and Bonito. The proposed accelerator maintains competitive throughput at 8 M Bases/sec, balancing processing speed and energy efficiency. Additionally, the architecture supports scalability up to 4096 states, making it highly adaptable for various sequencing applications. It’s hardware-optimized and low-power design makes it an ideal alternative to brute-force and software-based methods for real-time, mobile, and embedded DNA sequencing devices.
Real-time intrusion detection in heterogeneous Internet of Things (IoT) networks involves continuously monitoring diverse connected devices and communication protocols to promptly identify malicious activities or anomalies. Due to varied device capabilities, dynamic topologies, and resource constraints, these systems leverage lightweight AI-driven analytics, edge processing, and adaptive security models to ensure minimal latency. Effective detection enhances resilience, safeguards sensitive data, and maintains seamless IoT operations in mission-critical environments. We propose a stage-specific Recursive Sparse & Relevance-based Feature Selection (RS2FS) and a confidence-gated Support Vector Machine (SVM) → SVM → ANFIS cascade for real-time intrusion detection in heterogeneous IoT networks. RS2FS combines elastic-net screening, MI ∩ mRMR relevance, stability selection, and margin-aware recursive pruning to yield compact, non-redundant feature sets per cascade stage. The cascade accepts easy cases with calibrated SVMs and routes ambiguous, family-localized traffic to per-family ANFIS rules, providing interpretable subtype decisions. Evaluated on CICIoT2023 with scenario-held-out splits (5 × grouped CV), our model attains Macro-F1 = 0.962, Macro-AUC = 0.991, Balanced Accuracy = 0.963, MCC = 0.952, Brier = 0.038, and ECE = 0.012 at 6.3 ms CPU latency per window with a 7.8 MB footprint. Class-wise F1 shows consistent gains: Benign 0.991, DDoS 0.984, DoS 0.958, Recon 0.961, Web 0.937, Brute Force 0.951, Data Exfiltration 0.921, Botnet 0.942. Cascade behavior explains the speed-accuracy trade-off: 68% of windows are resolved at Stage-1 (F1 0.985, 3.38 ms), 22% at Stage-2 (F1 0.962, 7.73 ms), and only 10% escalate to ANFIS (F1 0.936, 23 ms). Against strong baselines, we improve Macro-F1 by + 1.9 pp over SVM-only (0.943), + 1.7 pp over XGBoost (0.945), and + 1.1 pp over a small 1D-CNN (0.951); bootstrap tests show significance (p < 0.01). Unlike existing IoT IDS approaches that rely on single-stage classifiers or one-time, global feature selection, our framework introduces two fundamental advances. First, the proposed RS2FS mechanism performs stage-specific, stability-aware, and margin-guided feature reduction, addressing the gaps of redundancy, volatility, and non-adaptiveness found in prior MI-, mRMR-, or L1-based selection methods. Second, the confidence-gated SVM → SVM → ANFIS cascade introduces a new routing paradigm where high-margin "easy" traffic is settled early, while only low-confidence, ambiguous windows are escalated to fuzzy reasoning overcoming the limitations of conventional hybrid SVM-ANFIS systems that apply the same classifier depth to all samples. Together with integrated open-set rejection and drift micro-adaptation, these contributions position the framework as a fundamentally new IDS architecture for heterogeneous IoT environments. The open-set guard achieves AUROC 0.981 and TPR@1%FPR 0.912 with 4.6% reject rate. Robustness holds under + 5% timestamp jitter (0.957), ± 10% packet-size noise (0.955), and 10% missing features (0.949). Interpretable ANFIS rules highlight payload-entropy, MQTT topic-depth, and DWT-energy interactions. Overall, the framework delivers accurate, calibrated, interpretable, and fast IDS suitable for deployment in modern IoT environments.
This paper presents an Echo State Network (ESN)-Driven joint decision strategy for a configurable security Non-Orthogonal Multiple Access (NOMA) optical communication system. To address security vulnerabilities, a configurable key iteration scheme is introduced, achieving temporal interleaving of chaotic models and keys. Furthermore, an ESN-Driven joint demodulation strategy is designed, which replaces conventional log-likelihood ratio (LLR) computations with ESN, leading to a reduction in computational overhead. Experimental validation over a 7-core fiber demonstrates that the ESN-Driven joint decision strategy maintains low computational cost while increasing Brute-Force Time (BFT) to 10151, thereby significantly enhancing efficiency and security of the system.
This study presents an efficient image encryption algorithm designed for secure data transmission in big data environments. The proposed method employs a cosine extended logistic chaotic map with a wider chaotic range and enhanced randomness. The map's dynamics are verified through bifurcation diagrams, Lyapunov exponents, Shannon entropy and Kolmogorov entropy. The encryption scheme incorporates two confusion and two diffusion phases using chaotic pixel permutation, controlled flipping, modulo arithmetic, MSB/LSB separation, and cross quadrant bitwise operations to achieve lightweight yet robust protection suitable for resource constrained systems. Experimental results on standard USC-SIPI and medical image datasets show near ideal entropy (~ 7.997), NPCR and UACI values close to 99.6094% and 33.4635%, and chi square and correlation values within secure limits, indicating strong resistance against statistical and differential attacks. With an estimated key space of 2318, the scheme surpasses brute force resilience standards and demonstrates an effective balance between security and computational efficiency for real time image transmission.
Metal-organic frameworks (MOFs) are prime candidate materials for gas adsorption and separation owing to their exceptional porosity and structural tunability. However, the nearly infinite chemical space and exponentially growing number of candidate structures pose insurmountable challenges to traditional experimental methods and brute-force computational screening. Data-driven machine learning (ML) offers a transformative solution for efficiently navigating this vast materials library. This review analyzes the current state of ML-based MOF screening, evaluates the limitations of mainstream MOF databases, and highlights how data authenticity and update frequency affect model reliability. The evolution of feature engineering─from manual geometric descriptors to automated representation learning using graph neural networks (GNNs) and molecular fingerprints─is also outlined. Furthermore, we discuss the specific applicability of advanced algorithmic frameworks, including deep learning, active learning, and transformers, to MOF screening tasks. Future development should focus on integrating high-fidelity experimental data with model interpretability to enable closed-loop autonomous discovery systems.
Internet-wide scanning is indispensable for security research and network measurement, yet its efficacy remains limited by significant visibility heterogeneity across global networks. Traditional centralized scanners suffer from single-point failures and offer a constrained perspective, while naive distributed approaches fail to intelligently leverage visibility variations, leading to redundant effort and incomplete coverage. This paper presents VistaScan, a novel distributed scanning system based on node visibility awareness, demonstrating that the visibility pattern among IP addresses is highly consistent within CIDR blocks, enabling a lightweight method for efficient and scalable quantification of per-node visibility. It first constructs a Visibility Matrix through efficient anchor probing, then employs a load-aware task allocation mechanism that assigns each block to the most capable node while filtering out entirely invisible blocks. Evaluation across global, regional, and challenging Special-Block tasks demonstrates that VistaScan consistently outperforms five baseline methods. It achieves near-optimal coverage (97.95%, 99.05%, and 97.58%, respectively), reduces probe volume by 64-93%, and shortens completion time by 13-19× compared to conventional centralized and distributed scanners. Furthermore, the visibility matrix derived from one port (TCP/80) effectively generalizes to other TCP ports (TCP/22, TCP/53), achieving coverages of 91.09% and 87.95%-preliminarily validating the practical generalizability of our approach. VistaScan provides both a highly efficient solution for Internet-scale distributed measurement and a new theoretical foundation based on visibility consistency, advancing the field from brute-force probing toward intelligent, low-overhead, and sustainable scanning practices.
The Profitable Tour Problem is a well-known NP-hard optimization challenge central to tourism planning, aiming to maximize collected profit while minimizing travel costs. While classical heuristics provide approximate solutions, they often struggle with finding globally optimal routes. This paper explores the application of near-term quantum computing to this problem. We propose a framework based on the Variational Quantum Eigensolver to find high-quality solutions for the Profitable Tour Problem. The core of our contribution is a novel methodology for constructing a constraint-aware variational ansatz that directly encodes the problem's hard constraints. This approach circumvents the need for large penalty terms in the Hamiltonian problem, which are often a source of optimization challenges. We validate our method through numerical simulations on a representative tourism scenario of up to 25 qubits. The results demonstrate the viability of the approach, achieving high solution accuracy consistent with brute-force enumeration for smaller instances. This work serves as a proof-of-concept for applying Variational Quantum Eigensolver to complex tourism optimization problems and provides a basis for future exploration on real quantum hardware.
Over half of those who quit smoking do so without formal assistance, yet the psychological processes supporting unassisted cessation remain little understood. Success is often attributed to willpower, an umbrella term that lacks explanatory precision and obscures the underlying tractable processes. Drawing on the Process Model of Self-Regulation and the Behavior Change Technique (BCT) Taxonomy, this study aimed to identify the concrete strategies that enable individuals to quit smoking unassisted, thereby clarifying what willpower might look like in practice. Thirty-two participants who had successfully quit smoking without formal support participated in semi-structured interviews. Inductive content analysis identified key challenges, while deductive coding mapped strategies addressing these challenges to the Process Model of Self-Regulation and the BCT Taxonomy. Participants' accounts reflected a diverse range of strategies, averaging seven distinct BCTs, spanning the Situation Selection and Modification, Attention Redeployment, and Cognitive Change stages from the Process Model. Common BCTs included avoiding environmental triggers, substituting smoking with alternative behaviors, and seeking social support. In contrast, Response Modulation (e.g. 'just say no') accounted for only 1% of the data. Unassisted quitters drew from a sophisticated repertoire of strategies that are actionable, teachable, and embedded within the individual's physical and social environment. The qualitative methodology used in this study offers an understanding of the lived experiences of self-quitters, potentially informing public health interventions that integrate individual and system-level approaches to behavior change that extend beyond brute-force willpower.
We describe AI agents as stochastic dynamical systems and frame the problem of learning to reason as in transductive inference: Rather than approximating the distribution of past data as in classical induction, the objective is to capture its algorithmic structure so as to reduce the time needed to solve new tasks. In this view, information from past experience serves not only to reduce a model's uncertainty, as in Shannon's classical theory, but to reduce the computational effort required to find solutions to unforeseen tasks. Working in the verifiable setting, where a checker or reward function is available, we establish three main results. First, we show that the optimal speed-up for a new task is tightly related to the algorithmic information it shares with the training data, yielding a theoretical justification for the power-law scaling empirically observed in reasoning models. Second, while the compression view of learning, rooted in Occam's Razor, favors simplicity, we show that transductive inference yields its greatest benefits precisely when the data-generating mechanism is most complex. Third, we identify a possible failure mode of naïve scaling: in the limit of unbounded model size and computing, models with access to a reward signal can behave as savants, brute-forcing solutions without acquiring transferable reasoning strategies. Accordingly, we argue that a critical quantity to optimize when scaling reasoning models is time, the role of which in learning has remained largely unexplored.
Ensuring the confidentiality of medical images during storage and transmission is a critical challenge in modern healthcare. This study proposes a novel hybrid image encryption framework that integrates a memristor-based chaotic system, DNA-inspired operations, Add-Rotate-Xor (ARX), and Triple Data Encryption Standard (3DES). A four-dimensional two-memristor chaotic circuit is analysed through phase portraits, Lyapunov exponents, and bifurcation diagrams to establish its suitability as a strong entropy source. Chaotic sequences generated from the system are digitized using a mod2 post-processing scheme and validated by NIST SP 800 − 22, FIPS 140-1, and Chi-square statistical tests, confirming high-quality randomness. The encryption framework combines chaotic diffusion and confusion, symbolic DNA crossover operations, ARX, and a 3DES whitening stage to provide multilayered security. Experimental validation on four medical image datasets—Bone Fracture, Breast, Retina, and Teeth—demonstrates that the scheme achieves near-ideal entropy values (~ 7.99), high NPCR (~99.6%), and UACI (~ 33%), while producing cipher images with noise-like characteristics and negligible structural similarity to the originals. Real-time implementation on the NVIDIA Jetson Nano and PYNQ-Z1 platform verifies the feasibility of the method for embedded medical applications. Comparative analysis indicates that the hybrid approach outperforms conventional chaotic or DNA-only schemes by leveraging the complementary strengths of chaos, bio-inspired computing, and classical cryptography. These findings confirm that the proposed framework offers a secure, efficient, and practical solution for protecting sensitive medical images against statistical, differential, and brute-force attacks.
Numerous computational approaches have been developed to infer cell state transition trajectories from snapshot single-cell data. Most approaches first require projecting high-dimensional data onto a low-dimensional representation; however, this can distort the dynamics of the system. Using epithelial-to-mesenchymal transition (EMT) as a test system, we show that both biology-guided low-dimensional representations and trajectory simulations in high-dimensional state space, not representations obtained with brute force dimensionality-reduction methods, reveal two broad paths of TGF-β induced EMT. The paths arise from the coupling between cell cycle and EMT at either the G1 or G2/M phase, contributing to cell-cycle related EMT heterogeneity. Subsequent multi-plex immunostaining studies confirmed the multiple predicted paths at the protein level. The present study highlights the heterogeneity of EMT paths, emphasizes that caution should be taken when inferring transition dynamics from snapshot single-cell data in two- or three-dimensional representations, and shows that incorporating dynamical information can improve prediction accuracy.
Expansion of diffusion MRI (dMRI) both into the realm of strong gradients and into accessible imaging with portable low-field devices brings about the challenge of gradient nonlinearities. Spatial variations of the diffusion gradients make diffusion weightings and directions non-uniform across the field of view, and deform perfect shells in the q $$ q $$ -space designed for isotropic directional coverage. Such imperfections hinder parameter estimation: Anisotropic shells hamper the deconvolution of the fiber orientation distribution function (fODF), while brute-force retraining of a nonlinear regressor for each unique set of directions and diffusion weightings is computationally inefficient. Here, we propose a protocol-independent parameter estimation (PIPE) method that enables fast parameter estimation for the most general case where each voxel is measured with a different protocol in q $$ q $$ -space. PIPE applies to any spherical convolution-based dMRI model, irrespective of its complexity, which makes it suitable both for white and gray matter in the brain or spinal cord, and for other tissues where fiber bundles have the same properties (fiber response) within a voxel, but are distributed with an arbitrary fODF. We also derive a parsimonious representation that isolates isotropic and anisotropic effects of gradient nonlinearities on multidimensional diffusion encodings. Applied to in vivo human MRI with linear tensor encoding on a high-performance gradient system, PIPE evaluates fiber response and fODF parameters for the whole brain in the presence of significant gradient nonlinearities in under 3 min. PIPE enables fast parameter estimation in the presence of arbitrary gradient nonlinearities, eliminating the need to arrange dMRI in shells or to retrain the estimator for different protocols in each voxel. PIPE applies to any model based on a convolution of a voxel-wise fiber response and fODF, and data from varying b $$ b $$ -values, diffusion/echo times, and other scan parameters.
As color images have become a cornerstone of information exchange in the digital age, ensuring their security is of paramount importance. With the traditional scrambling-diffusion structure, this paper proposes a novel color image encryption algorithm by integrating of an ∞ -shaped transformation with a closed-loop control mechanism. First, the three channel matrices are merged, and the elements in each row are linked into a closed loop for initial diffusion. Secend, the diffused image is subsequently restructured into a three-row matrix and scrambled using a unique ∞ -shaped transformation. Finally, column-wise closed-loop diffusion is applied to generate the cipher image. This algorithm not only achieves effective inter-channel pixel confusion through the ∞ -shaped transformation, but also performs additional diffusion and confusion under the closed-loop control model. Experimental results demonstrate the algorithm's excellent overall performance: the key space is as large as 2413, information entropy approaches the ideal value of 8 with increasing image size, and the algorithm exhibits high sensitivity, with NPCR and UACI exceeding 99.6% and 33.4%, respectively. Quantitative evaluation confirms that the proposed algorithm offers strong robustness against differential, statistical, and brute-force attacks.
Biometric encryption integrates physiological traits with cryptographic operations to improve authentication security. Retinal vasculature is particularly attractive due to its internal protection, permanence, and high inter-subject variability. we present a revised and rigorously justified multidimensional retinal encryption framework that generates three independent keys—RDDM (Retinal Diagonal Distance Metric), ROTD (Radial Origin-Terminus Distance), and DRID (Diagonal–Radial Intersection Distance)—from a single retinal vessel map. This framework is intended as a biometric-driven key generation and strengthening module to enhance user authentication, rather than a standalone standard encryption algorithm. It operates under a threat model focused on resisting brute-force key guessing in controlled biometric contexts, but not advanced attacks like quantum cryptanalysis or side-channel exploitation. Retinal images undergo preprocessing (CLAHE, vessel segmentation, skeletonization, endpoint detection) to extract stable endpoints. These endpoints produce distance measures that are normalized and combined into polyalphabetic key streams. We provide stepwise derivations of the encryption E(x) and decryption D(y) equations, explicitly justify mod 124 as the symbol table size used in implementation, and include a detailed cryptanalytic evaluation (entropy, NIST SP800-22 randomness tests, Hamming distance, collision analysis, noise sensitivity, and Full-Space Key Guessing (FSKG) calculations). Experimental results on three retinal samples (n_vessels = 27,41,105) show substantially increased FSKG times and near-maximal key entropy relative to single-key baselines. Limitations and sensitivity to image quality are discussed.