arXiv subjects

Danyang Li

Publications and source records attributed to Danyang Li.

30 records · Page 2Linked to original sources

arXiv2025

Interpretable Imitation Learning via Generative Adversarial STL Inference and Control

Imitation learning methods have demonstrated considerable success in teaching autonomous systems complex tasks through expert demonstrations. However, a limitation of these methods is their lack of interpretability, particularly in understanding the specific task the learning agent aims to accomplish. In this paper, we propose a novel imitation learning method that combines Signal Temporal Logic (STL) inference and control synthesis, enabling the explicit representation of the task as an STL formula. This approach not only provides a clear understanding of the task but also supports the integration of human knowledge and allows for adaptation to out-of-distribution scenarios by manually adjusting the STL formulas and fine-tuning the policy. We employ a Generative Adversarial Network (GAN)-inspired approach to train both the inference and policy networks, effectively narrowing the gap between expert and learned policies. The efficiency of our algorithm is demonstrated through simulations, showcasing its practical applicability and adaptability.

cs.LG↗

arXiv2025

SpecOffload: Unlocking Latent GPU Capacity for LLM Inference on Resource-Constrained Devices

Efficient LLM inference on resource-constrained devices presents significant challenges in compute and memory utilization. Due to limited GPU memory, existing systems offload model weights to CPU memory, incurring substantial I/O overhead between the CPU and GPU. This leads to two major inefficiencies: (1) GPU cores are underutilized, often remaining idle while waiting for data to be loaded; and (2) GPU memory has low impact on performance, as reducing its capacity has minimal effect on overall throughput.In this paper, we propose SpecOffload, a high-throughput inference engine that embeds speculative decoding into offloading. Our key idea is to unlock latent GPU resources for storing and executing a draft model used for speculative decoding, thus accelerating inference at near-zero additional cost. To support this, we carefully orchestrate the interleaved execution of target and draft models in speculative decoding within the offloading pipeline, and propose a planner to manage tensor placement and select optimal parameters. Compared to the best baseline, SpecOffload improves GPU core utilization by 4.49x and boosts inference throughput by 2.54x. Our code is available at https://github.com/MobiSense/SpecOffload-public .

cs.LG↗

arXiv2025

A Simple Detector with Frame Dynamics is a Strong Tracker

Infrared object tracking plays a crucial role in Anti-Unmanned Aerial Vehicle (Anti-UAV) applications. Existing trackers often depend on cropped template regions and have limited motion modeling capabilities, which pose challenges when dealing with tiny targets. To address this, we propose a simple yet effective infrared tiny-object tracker that enhances tracking performance by integrating global detection and motion-aware learning with temporal priors. Our method is based on object detection and achieves significant improvements through two key innovations. First, we introduce frame dynamics, leveraging frame difference and optical flow to encode both prior target features and motion characteristics at the input level, enabling the model to better distinguish the target from background clutter. Second, we propose a trajectory constraint filtering strategy in the post-processing stage, utilizing spatio-temporal priors to suppress false positives and enhance tracking robustness. Extensive experiments show that our method consistently outperforms existing approaches across multiple metrics in challenging infrared UAV tracking scenarios. Notably, we achieve state-of-the-art performance in the 4th Anti-UAV Challenge, securing 1st place in Track 1 and 2nd place in Track 2.

cs.CV↗

arXiv2025

Unprecedented Continuum Time Delay Variation Phenomenon in an Active Supermassive Black Hole

Resolving the accretion disks and broad line regions (BLRs) of active galactic nuclei (AGNs) can probe the physics behind supermassive black holes (SMBHs) fueling and weigh SMBHs. With time-domain observations, the reverberation mapping (RM) technique measures time delays between different light curves, probing the AGN inner structures that are otherwise often too compact to resolve spatially with current facilities. Theoretically, the SMBH accretion disk structure does not evolve over decades. Here we report the significant variations in the continuum time delays of NGC 4151. In the high-flux state, our high-cadence ($\sim 2$ days) spectroscopy reveals that continuum time delays are $3.8^{+1.8}_{-1.0}$ times larger than those in the low-flux state and $14.9\pm 2.0$ times longer than the classical standard thin disk prediction. Notably, the continuum time delays can be comparable with the time delay between H$β$ and the $5100\ \mathrmÅ$ continuum, and the latter is commonly used to calculate the BLR sizes. Hence, the BLR sizes are underestimated if the continuum time delays are not considered, which introduces $\sim 0.3$ dex systematic uncertainties on RM SMBH masses. Our findings underscore that simultaneous continuum and BLR RMs are vital for better deciphering the SMBH accretion and mass function.

astro-ph.HE↗

arXiv2025

Enhancing the Accuracy and Comprehensibility in Architectural Tactics Detection via Small Model-Augmented Prompt Engineering

Architectural tactics (ATs), as the concrete implementation of architectural decisions in code, address non-functional requirements of software systems. Due to the implicit nature of architectural knowledge in code implementation, developers may risk inadvertently altering or removing these tactics during code modifications or optimizations. Such unintended changes can trigger architectural erosion, gradually undermining the system's original design. While many researchers have proposed machine learning-based methods to improve the accuracy of detecting ATs in code, the black-box nature and the required architectural domain knowledge pose significant challenges for developers in verifying the results. Effective verification requires not only accurate detection results but also interpretable explanations that enhance their comprehensibility. However, this is a critical gap in current research. Large language models (LLMs) can generate easily interpretable ATs detection comments if they have domain knowledge. Fine-tuning LLMs to acquire domain knowledge faces challenges such as catastrophic forgetting and hardware constraints. Thus, we propose Prmt4TD, a small model-augmented prompting framework to enhance the accuracy and comprehensibility of ATs detection. Combining fine-tuned small models with In-Context Learning can also reduce fine-tuning costs while equipping the LLM with additional domain knowledge. Prmt4TD can leverage the remarkable processing and reasoning capabilities of LLMs to generate easily interpretable ATs detection results. Our evaluation results demonstrate that Prmt4TD achieves accuracy (\emph{F1-score}) improvement of 13\%-23\% on the ATs balanced dataset and enhances the comprehensibility of the detection results.

cs.SE↗

arXiv2024

Data Interpreter: An LLM Agent For Data Science

Large Language Model (LLM)-based agents have shown effectiveness across many applications. However, their use in data science scenarios requiring solving long-term interconnected tasks, dynamic data adjustments and domain expertise remains challenging. Previous approaches primarily focus on individual tasks, making it difficult to assess the complete data science workflow. Moreover, they struggle to handle real-time changes in intermediate data and fail to adapt dynamically to evolving task dependencies inherent to data science problems. In this paper, we present Data Interpreter, an LLM-based agent designed to automatically solve various data science problems end-to-end. Our Data Interpreter incorporates two key modules: 1) Hierarchical Graph Modeling, which breaks down complex problems into manageable subproblems, enabling dynamic node generation and graph optimization; and 2) Programmable Node Generation, a technique that refines and verifies each subproblem to iteratively improve code generation results and robustness. Extensive experiments consistently demonstrate the superiority of Data Interpreter. On InfiAgent-DABench, it achieves a 25% performance boost, raising accuracy from 75.9% to 94.9%. For machine learning and open-ended tasks, it improves performance from 88% to 95%, and from 60% to 97%, respectively. Moreover, on the MATH dataset, Data Interpreter achieves remarkable performance with a 26% improvement compared to state-of-the-art baselines. The code is available at https://github.com/geekan/MetaGPT.

cs.AI↗

arXiv2024

Multi-class Temporal Logic Neural Networks

Time-series data can represent the behaviors of autonomous systems, such as drones and self-driving cars. The task of binary and multi-class classification for time-series data has become a prominent area of research. Neural networks represent a popular approach to classifying data; However, they lack interpretability, which poses a significant challenge in extracting meaningful information from them. Signal Temporal Logic (STL) is a formalism that describes the properties of timed behaviors. We propose a method that combines all of the above: neural networks that represent STL specifications for multi-class classification of time-series data. We offer two key contributions: 1) We introduce a notion of margin for multi-class classification, and 2) we introduce STL-based attributes for enhancing the interpretability of the results. We evaluate our method on two datasets and compare it with state-of-the-art baselines.

stat.ML↗

arXiv2024

TLINet: Differentiable Neural Network Temporal Logic Inference

There has been a growing interest in extracting formal descriptions of the system behaviors from data. Signal Temporal Logic (STL) is an expressive formal language used to describe spatial-temporal properties with interpretability. This paper introduces TLINet, a neural-symbolic framework for learning STL formulas. The computation in TLINet is differentiable, enabling the usage of off-the-shelf gradient-based tools during the learning process. In contrast to existing approaches, we introduce approximation methods for max operator designed specifically for temporal logic-based gradient techniques, ensuring the correctness of STL satisfaction evaluation. Our framework not only learns the structure but also the parameters of STL formulas, allowing flexible combinations of operators and various logical structures. We validate TLINet against state-of-the-art baselines, demonstrating that our approach outperforms these baselines in terms of interpretability, compactness, rich expressibility, and computational efficiency.

cs.LO↗

arXiv2024

Room temperature single-photon terahertz detection with thermal Rydberg atoms

Single-photon terahertz (THz) detection is one of the most demanding technology for a variety of fields and could lead to many breakthroughs. Although its significant progress has been made in the last two decades, operating it at room temperature still remains a great challenge. Here, we demonstrate, for the first time, the room temperature THz detector at single-photon levels based on nonlinear wave mixing in thermal Rydberg atomic vapor. The low-energy THz photons are coherently upconverted to the high-energy optical photons via a nondegenerate Rydberg state involved six-wave-mixing process, and therefore, the single-photon THz detection is achieved by a conventional optical single-photon counting module. The noise equivalent power of such a detector is reached to be 9.5*10^-19 W/Hz^1/2, which is more than four orders of magnitude lower than the state-of-the-art room temperature THz detectors. The optimum quantum efficiency of the whole wave-mixing process is about 4.3% with 40.6 dB dynamic range, and the maximum conversion bandwidth is 172 MHz, which is all-optically controllable. The developed fast and continuous-wave single-photon THz detector at room temperature operation has a great potential to be portable and chip-scale, and could be revolutionary for a wide range of applications in remote sensing, wireless communication, biomedical diagnostics, and quantum optics.

quant-ph↗

arXiv2024

A reverberation mapping study of a highly variable AGN 6dFGS gJ022550.0-060145

We use LCOGT observations (MJD $59434-59600$) with a total exposure time of $\simeq 50$ hours and a median cadence of $0.5$ days to measure the inter-band time delays (with respect to $u$) in the $g$, $r$, and $i$ continua of a highly variable AGN, 6dFGS gJ022550.0-060145. We also calculate the expected time delays of the X-ray reprocessing of a static Shakura \& Sunyaev disk (SSD) according to the sources' luminosity and virial black-hole mass; the two parameters are measured from the optical spectrum of our spectroscopic observation via the Lijiang \SI{2.4}{\meter} telescope. It is found that the ratio of the measured time delays to the predicted ones is $2.6_{-1.3}^{+1.3}$. With optical light curves (MJD $53650-59880$) from our new LCOGT and archival ZTF, Pan-SATRRS, CSS, and ATLAS observations, and infrared (IR) WISE data (MJD $55214-59055$), we also measured time delays between WISE $W1$/$W2$ and the optical emission. $W1$ and $W2$ have time delays (with respect to V), $9.6^{+2.9}_{-1.6}\times 10^2$ days and $1.18^{+0.13}_{-0.10}\times 10^3$ days in the rest-frame, respectively; hence, the dusty torus of 6dFGS gJ022550.0-060145 should be compact. The time delays of $W1$ and $W2$ bands are higher than the dusty torus size-luminosity relationship of~\cite{Lyu2019}. By comparing the IR and optical variability amplitude, we find that the dust covering factors of $W1$ and $W2$ emission regions are 0.7 and 0.6, respectively. Future broad emission-line reverberation mapping of this target and the results of this work enable us to determine the sizes of the AGN main components simultaneously.

astro-ph.HE↗

arXiv2023

Learning Signal Temporal Logic through Neural Network for Interpretable Classification

Machine learning techniques using neural networks have achieved promising success for time-series data classification. However, the models that they produce are challenging to verify and interpret. In this paper, we propose an explainable neural-symbolic framework for the classification of time-series behaviors. In particular, we use an expressive formal language, namely Signal Temporal Logic (STL), to constrain the search of the computation graph for a neural network. We design a novel time function and sparse softmax function to improve the soundness and precision of the neural-STL framework. As a result, we can efficiently learn a compact STL formula for the classification of time-series data through off-the-shelf gradient-based tools. We demonstrate the computational efficiency, compactness, and interpretability of the proposed method through driving scenarios and naval surveillance case studies, compared with state-of-the-art baselines.

cs.FL↗

arXiv2023

Soft isogeometric analysis of the Bound States of a Quantum Three-Body Problem in 1D

The study of quantum three-body problems has been centered on low-energy states that rely on accurate numerical approximation. Recently, isogeometric analysis (IGA) has been adopted to solve the problem as an alternative but more robust (with respect to atom mass ratios) method that outperforms the classical Born-Oppenheimer (BO) approximation. In this paper, we focus on the performance of IGA and apply the recently-developed softIGA to reduce the spectral errors of the low-energy bound states. The main idea is to add high-order derivative-jump terms with a penalty parameter to the IGA bilinear forms. With an optimal choice of the penalty parameter, we observe eigenvalue error superconvergence. We focus on linear (finite elements) and quadratic elements and demonstrate the outperformance of softIGA over IGA through a variety of examples including both two- and three-body problems in 1D.

math.NA↗