arXiv subjects

Qifeng Li

Publications and source records attributed to Qifeng Li.

At least 37 records · Page 2Linked to original sources

arXiv2025

Real-time Optimization for Wind-to-H2 Driven Critical Infrastructures Based on Active Constraints Identification and Integer Variables Prediction

This paper proposes a concept of wind-to-hydrogen-driven critical infrastructure (W2H-CI) as an engineering solution for decarbonizing the power generation sector where it utilizes wind power to produce hydrogen through electrolysis and combines it with the carbon captured from fossil fuel power plants. First, a convex mathematical model of W2H-CI is developed. Then, an optimization model for optimal operation of W2H-CI, which is a large-scale mixed-integer convex program (MICP), is proposed. Moreover, we propose to solve this problem in real-time in order to hedge against the uncertainty of wind power. For this purpose, a novel solution method based on active constraints identification and integer variable prediction is introduced. This method can solve MICP problems very fast since it uses historical optimization data to predict the values of binary variables and a limited number of constraints which most likely contain all active constraints. We validate the effectiveness of the proposed fast solution method using two W2H-CI case studies.

eess.SY↗

arXiv2024

Automatically Planning Optimal Parallel Strategy for Large Language Models

The number of parameters in large-scale language models based on transformers is gradually increasing, and the scale of computing clusters is also growing. The technology of quickly mobilizing large amounts of computing resources for parallel computing is becoming increasingly important. In this paper, we propose an automatic parallel algorithm that automatically plans the parallel strategy with maximum throughput based on model and hardware information. By decoupling the training time into computation, communication, and overlap, we established a training duration simulation model. Based on this simulation model, we prune the parallel solution space to shorten the search time required. The multi-node experiment results show that the algorithm can estimate the parallel training duration in real time with an average accuracy of 96%. In our test, the recommendation strategy provided by the algorithm is always globally optimal.

cs.AI↗

arXiv2024

Field-free current-induced magnetization switching of a room temperature van der Waals magnet for neuromorphic computing

Spin orbit torque (SOT) has become a promising approach to efficiently manipulate the magnetization switching in spintronic devices. As a main factor to impact the device performance, the high quality interface is essentially desired, which can be readily acquired by using the two-dimensional (2D) van der Waals (vdW) materials. Recently, a 2D ferromagnetic material Fe3GaTe2 has been discovered to possess the above-room-temperature Curie temperature and strong perpendicular magnetic anisotropy (PMA), providing an excellent candidate to build spintronic devices. On the other hand, an external magnetic field is necessary for the SOT-driven deterministic switching of perpendicular magnetization, which has become a block for the real applications. Here, we realize the field-free SOT switching of Fe3GaTe2 at room temperature based on the Fe3GaTe2/MnPt heterostructure. In addition, inspired by the superiority of 2D materials in 3D heterogeneous integration, we explore the potential of our device in the computing in memory (CIM). With the application of the current pulses, the gradual switching of our device at zero field imitates the function of artificial synapse in the convolutional neural network (CNN), achieving a high accuracy (~92.8%) pattern recognition. Our work proposes a feasible solution for field-free SOT switching in 2D vdW spintronic devices, which paves the way for applications in magnetic memory and neuromorphic computing.

physics.app-ph↗

arXiv2024

Bench2Drive: Towards Multi-Ability Benchmarking of Closed-Loop End-To-End Autonomous Driving

In an era marked by the rapid scaling of foundation models, autonomous driving technologies are approaching a transformative threshold where end-to-end autonomous driving (E2E-AD) emerges due to its potential of scaling up in the data-driven manner. However, existing E2E-AD methods are mostly evaluated under the open-loop log-replay manner with L2 errors and collision rate as metrics (e.g., in nuScenes), which could not fully reflect the driving performance of algorithms as recently acknowledged in the community. For those E2E-AD methods evaluated under the closed-loop protocol, they are tested in fixed routes (e.g., Town05Long and Longest6 in CARLA) with the driving score as metrics, which is known for high variance due to the unsmoothed metric function and large randomness in the long route. Besides, these methods usually collect their own data for training, which makes algorithm-level fair comparison infeasible. To fulfill the paramount need of comprehensive, realistic, and fair testing environments for Full Self-Driving (FSD), we present Bench2Drive, the first benchmark for evaluating E2E-AD systems' multiple abilities in a closed-loop manner. Bench2Drive's official training data consists of 2 million fully annotated frames, collected from 13638 short clips uniformly distributed under 44 interactive scenarios (cut-in, overtaking, detour, etc), 23 weathers (sunny, foggy, rainy, etc), and 12 towns (urban, village, university, etc) in CARLA v2. Its evaluation protocol requires E2E-AD models to pass 44 interactive scenarios under different locations and weathers which sums up to 220 routes and thus provides a comprehensive and disentangled assessment about their driving capability under different situations. We implement state-of-the-art E2E-AD models and evaluate them in Bench2Drive, providing insights regarding current status and future directions.

cs.RO↗

arXiv2024

An Unsupervised Dialogue Topic Segmentation Model Based on Utterance Rewriting

Dialogue topic segmentation plays a crucial role in various types of dialogue modeling tasks. The state-of-the-art unsupervised DTS methods learn topic-aware discourse representations from conversation data through adjacent discourse matching and pseudo segmentation to further mine useful clues in unlabeled conversational relations. However, in multi-round dialogs, discourses often have co-references or omissions, leading to the fact that direct use of these discourses for representation learning may negatively affect the semantic similarity computation in the neighboring discourse matching task. In order to fully utilize the useful cues in conversational relations, this study proposes a novel unsupervised dialog topic segmentation method that combines the Utterance Rewriting (UR) technique with an unsupervised learning algorithm to efficiently utilize the useful cues in unlabeled dialogs by rewriting the dialogs in order to recover the co-referents and omitted words. Compared with existing unsupervised models, the proposed Discourse Rewriting Topic Segmentation Model (UR-DTS) significantly improves the accuracy of topic segmentation. The main finding is that the performance on DialSeg711 improves by about 6% in terms of absolute error score and WD, achieving 11.42% in terms of absolute error score and 12.97% in terms of WD. on Doc2Dial the absolute error score and WD improves by about 3% and 2%, respectively, resulting in SOTA reaching 35.17% in terms of absolute error score and 38.49% in terms of WD. This shows that the model is very effective in capturing the nuances of conversational topics, as well as the usefulness and challenges of utilizing unlabeled conversations.

cs.CL↗

arXiv2024

Characteristic conic connections and torsion-free principal connections

We study the relation between torsion tensors of principal connections on G-structures and characteristic conic connections on associated cone structures. We formulate sufficient conditions under which the existence of a characteristic conic connection implies the existence of a torsion-free principal connection. We verify these conditions for adjoint varieties of simple Lie algebras, excluding those of type $\textsf{A}_{\ell \neq 2}$ or $\textsf{C}_{\ell}$. As an application, we give a complete classification of the germs of minimal rational curves whose VMRT at a general point is such an adjoint variety: nontrivial ones come from lines on hyperplane sections of certain Grassmannians or minimal rational curves on wonderful group compactifications.

math.DG↗

arXiv2024

Think2Drive: Efficient Reinforcement Learning by Thinking in Latent World Model for Quasi-Realistic Autonomous Driving (in CARLA-v2)

Real-world autonomous driving (AD) especially urban driving involves many corner cases. The lately released AD simulator CARLA v2 adds 39 common events in the driving scene, and provide more quasi-realistic testbed compared to CARLA v1. It poses new challenge to the community and so far no literature has reported any success on the new scenarios in V2 as existing works mostly have to rely on specific rules for planning yet they cannot cover the more complex cases in CARLA v2. In this work, we take the initiative of directly training a planner and the hope is to handle the corner cases flexibly and effectively, which we believe is also the future of AD. To our best knowledge, we develop the first model-based RL method named Think2Drive for AD, with a world model to learn the transitions of the environment, and then it acts as a neural simulator to train the planner. This paradigm significantly boosts the training efficiency due to the low dimensional state space and parallel computing of tensors in the world model. As a result, Think2Drive is able to run in an expert-level proficiency in CARLA v2 within 3 days of training on a single A6000 GPU, and to our best knowledge, so far there is no reported success (100\% route completion)on CARLA v2. We also propose CornerCase-Repository, a benchmark that supports the evaluation of driving models by scenarios. Additionally, we propose a new and balanced metric to evaluate the performance by route completion, infraction number, and scenario density, so that the driving score could give more information about the actual driving performance.

cs.RO↗

arXiv2024

Raw Text is All you Need: Knowledge-intensive Multi-turn Instruction Tuning for Large Language Model

Instruction tuning as an effective technique aligns the outputs of large language models (LLMs) with human preference. But how to generate the seasonal multi-turn dialogues from raw documents for instruction tuning still requires further exploration. In this paper, we present a novel framework named R2S that leverages the CoD-Chain of Dialogue logic to guide large language models (LLMs) in generating knowledge-intensive multi-turn dialogues for instruction tuning. By integrating raw documents from both open-source datasets and domain-specific web-crawled documents into a benchmark K-BENCH, we cover diverse areas such as Wikipedia (English), Science (Chinese), and Artifacts (Chinese). Our approach first decides the logic flow of the current dialogue and then prompts LLMs to produce key phrases for sourcing relevant response content. This methodology enables the creation of the G I NSTRUCT instruction dataset, retaining raw document knowledge within dialoguestyle interactions. Utilizing this dataset, we fine-tune GLLM, a model designed to transform raw documents into structured multi-turn dialogues, thereby injecting comprehensive domain knowledge into the SFT model for enhanced instruction tuning. This work signifies a stride towards refining the adaptability and effectiveness of LLMs in processing and generating more accurate, contextually nuanced responses across various fields.

cs.CL↗

arXiv2024

End-to-End Streaming Video Temporal Action Segmentation with Reinforce Learning

The streaming temporal action segmentation (STAS) task, a supplementary task of temporal action segmentation (TAS), has not received adequate attention in the field of video understanding. Existing TAS methods are constrained to offline scenarios due to their heavy reliance on multimodal features and complete contextual information. The STAS task requires the model to classify each frame of the entire untrimmed video sequence clip by clip in time, thereby extending the applicability of TAS methods to online scenarios. However, directly applying existing TAS methods to SATS tasks results in significantly poor segmentation outcomes. In this paper, we thoroughly analyze the fundamental differences between STAS tasks and TAS tasks, attributing the severe performance degradation when transferring models to model bias and optimization dilemmas. We introduce an end-to-end streaming video temporal action segmentation model with reinforcement learning (SVTAS-RL). The end-to-end modeling method mitigates the modeling bias introduced by the change in task nature and enhances the feasibility of online solutions. Reinforcement learning is utilized to alleviate the optimization dilemma. Through extensive experiments, the SVTAS-RL model significantly outperforms existing STAS models and achieves competitive performance to the state-of-the-art TAS model on multiple datasets under the same evaluation criteria, demonstrating notable advantages on the ultra-long video dataset EGTEA. Code is available at https://github.com/Thinksky5124/SVTAS.

cs.CV↗

arXiv2023

Micro Energy-Water-Hydrogen Nexus: Data-driven Real-time Optimal Operation

This paper extends a new concept of energy-water-hydrogen (EWH) nexus, which was recently developed as a solution for reducing carbon emissions from the generation side of power systems, to the distribution side. Under the concept of distribution-level EWH (micro EWH) nexus, renewable energy sources (RES) are utilized to meet the energy needs of a small community. To avoid the uncertainty caused by RESs, this paper aims to investigate the real-time optimal operation of the micro EWH nexus which is however a challenging optimization problem. First, such a large-scale mixed-integer nonlinear programming problem is relaxed into a mixed-integer convex program (MICP) by leveraging the effective convex-hull relaxation technique. Second, a fast data-driven solution method based on active constraint and integer variable prediction is presented, which can solve the MICP problem very fast since it utilizes historical optimization data to quickly predict binary variable values and a limited set of active constraints.

eess.SY↗

arXiv2023

Economic Viability of the Energy-Water-Hydrogen Nexus for Power System Decarbonization

This paper aims to evaluate the economic viability of the energy-water-hydrogen (EWH) nexus as a new solution for reducing carbon emissions from power systems. The urgency around climate change emphasizes the pressing need to mitigate carbon emissions, especially from the electricity sector, which accounts for a significant portion of total emissions in the US. In response, incorporating more renewable energy sources (RESs) and green hydrogen, created through water electrolysis and RES, stands out as a crucial strategy to combat climate challenges. We delve into various aspects of the EWH nexus, including carbon emissions from different power plants, capturing these emissions, and potential options for their reuse or storage. This paper involves modeling different sections of the EWH nexus and conducting an economic analysis across scenarios in power plants to determine optimal water supply methods, suitable chemical products for carbon reuse, and an appropriate carbon emission penalty to encourage emission reduction through the EWH nexus. The results indicate that reusing captured carbon emissions emerges as the most beneficial option across all power plant types. This finding underscores the potential of carbon reuse as a pivotal strategy within the EWH nexus framework for addressing carbon emissions.

eess.SY↗

arXiv2023

Rigidity of projective symmetric manifolds of Picard number 1 associated to composition algebras

To each complex composition algebra $\mathbb{A}$, there associates a projective symmetric manifold $X(\mathbb{A})$ of Picard number one, which is just a smooth hyperplane section of the following varieties ${\rm Lag}(3,6), {\rm Gr}(3,6), \mathbb{S}_6, E_7/P_7.$ In this paper, it is proven that these varieties are rigid, namely for any smooth family of projective manifolds over a connected base, if one fiber is isomorphic to $X(\mathbb{A})$, then every fiber is isomorphic to $X(\mathbb{A})$.

math.AG↗

arXiv2023

Rigidity of wonderful group compactifications under Fano deformations

For a complex connected semisimple linear algebraic group $G$ of adjoint type and of rank $n$, De Concini and Procesi constructed its wonderful compactification $\bar{G}$, which is a smooth Fano $G \times G$-variety of Picard number $n$ enjoying many interesting properties. In this paper, it is shown that the wonderful compactification $\bar{G}$ is rigid under Fano deformation. Namely, for any regular family of Fano manifolds over a connected base, if one fiber is isomorphic to $\bar{G}$, then so are all other fibers. This answers a question raised by Bien and Brion in their work on the local rigidity of wonderful varieties.

math.AG↗

arXiv2023

Real-time Optimization for Wind-to-H2 Driven Critical Infrastructures: High-fidelity Active Constraints and Integer Variables Prediction Enhanced by Feature Space Expansion

This paper focuses on developing a real-time optimal operation model for a new engineering system, wind-to-hydrogen-driven low-carbon critical infrastructure (W2H-LCCI), that utilizes wind power to generate hydrogen through electrolysis and combines it with carbon capture to reduce carbon emissions from the power sector. First, a convex mathematical model for W2H-LCCI is proposed, and then optimization models for its real-time decision-making are developed, which are mixed-integer convex programs (MICPs). Furthermore, since this large-scale MICP problem must be solved in real-time, a fast solution method based on active constraint and integer variable prediction (ACIVP) is presented. ACIVP method predicts the binary variable values and the set of limited-number constraints, which most likely contain all of the active constraints, based on historical optimization data. It results in only a small-scale continuous convex optimization problem needing to be solved by optimization solvers for W2H-LCCI real-time optimal operation. To increase the accuracy of the ACIVP method, feature space expansion (FSE) is employed, and a multi-stage ACIVP-FSE method is proposed. The effects of stage design and stage ordering on ACIVP-FSE performance are also discussed. We validate the effectiveness of the developed system and solution method using two water-energy nexus case studies.

eess.SY↗

arXiv2023

Stochastic Economic Dispatch Considering Demand Response and Endogenous Uncertainty

This paper considers endogenous uncertainty (EnU) in the stochastic economic dispatch (SED) problem, where the endogenous uncertainty means decision dependent uncertainty. In this problem, demand response (DR) commitment is the source of the EnU. Nevertheless, EnU is not well considered in existing literature. Our first contribution is to build up an optimization model of DR-involved SED under EnU (SED-DR-EnU). This is a computational challenging problem due to the EnU. Our second contribution is introducing a coupled learning enabled optimization algorithm which can effectively solve the proposed SED-DR-EnU problem. This strategy is tested on the IEEE 14 bus, and IEEE 39 bus systems, and the results showed the importance of considering EnU in the DR-involved SED problem.

math.OC↗

arXiv2023

A Novel Decentralized Algorithm for Coordinating the Optimal Power and Traffic Flows with EVs based on Variable Inner Loop Selection

The electric power distribution network (PDN) and the transportation network (TN) are generally operated/coordinated by different entities. However, they are coupled with each other due to electric vehicle charging stations (EVCSs). This paper proposes to coordinate the operation of the two systems via a fully decentralized framework where the PDN and TN operators solve their own operation problems by sharing only limited information. Nevertheless, the operation problems generally are in mixed-integer programming (MIP) form. To the best of our knowledge, the most existing decentralized/distributed optimization algorithms, such as the alternating direction method of multipliers (ADMM), are not always guaranteed to converge for such MIP problems. Therefore, a novel fully decentralized optimization algorithm is proposed, whose contributions include: 1) it is applicable to MIP problems with convergence and optimality guaranteed for mild assumptions, 2) it only requires limited information exchange between PDN and TN operators, which will help preserve the privacy of the two systems and reduce the investment in building communication channels, and 3) it is fully decentralized so that all the computations are carried out by PDN operator and TN coordinator only. Simulations on the test cases show the efficiency and efficacy of the proposed framework and algorithm.

eess.SY↗

arXiv2023

Lines on holomorphic contact manifolds and a generalization of $(2,3,5)$-distributions to higher dimensions

Since the celebrated work by Cartan, distributions with \nobreak{small} growth vector $(2,3,5)$ have been studied extensively. In the holomorphic setting, there is a natural correspondence between holomorphic $(2,3,5)$-distributions and nondegenerate lines on holomorphic contact manifolds of dimension 5. We generalize this correspondence to higher dimensions by studying nondegenerate lines on holomorphic contact manifolds and the corresponding class of distributions of small growth vector $(2m, 3m, 3m+2)$ for any positive integer $m$.

math.DG↗

arXiv2022

Recognizing the ${\rm G}_2$-horospherical manifold of Picard number 1 by varieties of minimal rational tangents

Pasquier and Perrin discovered that the ${\rm G}_2$-horospherical manifold ${\bf X}$ of Picard number 1 can be realized as a smooth specialization of the rational homogeneous space parameterizing the lines on the 5-dimensional hyperquadric, in other words, it can be deformed nontrivially to the rational homogeneous space. We show that ${\bf X}$ is the only smooth projective variety with this property. This is obtained as a consequence of our main result that ${\bf X}$ can be recognized by its VMRT, namely, a Fano manifold of Picard number 1 is biregular to ${\bf X}$ if and only if its VMRT at a general point is projectively isomorphic to that of ${\bf X}$. We employ the method the authors developed to solve the corresponding problem for symplectic Grassmannians, which constructs a flat Cartan connection in a neighborhood of a general minimal rational curve. In adapting this method to ${\bf X}$, we need an intricate study of the positivity/negativity of vector bundles with respect to a family of rational curves, which is subtler than the case of symplectic Grassmannians because of the nature of the differential geometric structure on ${\bf X}$ arising from VMRT.

math.AG↗