YHYifan He

Yifan He

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Yifan He has a total of 30 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • A Q-Learning-Based Hyper-Heuristic for Capacitated Electric Vehicle Routing Problem

    by Jian Lin, Xintao Wang, Rui Niu, Yifan He

    The capacitated electric vehicle routing problem (CEVRP) requires a set of shortest delivery routes for capacitated electric vehicles from a depot to multiple customer locations. In this paper, a Q-learning-based hyper-heuristic (QHH) is proposed to solve the CEVRP. First, a novel initialization method is proposed to generate capacity-feasible solutions with potentially shorter routes. Second, a repair method is devised to insert proper charging stations into routes, generate feasible solutions, optimize visits to charging stations, and subsequently improve the solutions. Then, a diverse set of six common low-level heuristics (LLHs) and several local search methods are employed as fundamental components of the QHH algorithm. Meanwhile, Q-learning serves as a guiding high-level heuristic, dynamically selecting the appropriate LLHs in response to the evolutionary process. Additionally, the Design of Experiments methodology is employed to investigate the performance of various parameter settings. Finally, the proposed QHH is evaluated on a large set of 60 CEVRP instances, demonstrating its superiority over four state-of-the-art algorithms.


  • A Reinforcement Learning-Based Population Hyper-Heuristic for Energy-Efficient Cloud Workflow Scheduling Problem

    by Yuanzhuang Li, Yifan He, Jian Lin, Zhan Xu, Shiyu Zhang

    As cloud computing undergoes rapid development, cloud workflow scheduling has gained significant attention as a critical area of research. However, the substantial energy consumption of cloud data centers in handling diverse workflow applications highlights the need for effective energy optimization strategies to promote green computing. This research investigates an energy-efficient cloud workflow scheduling problem that explores the inherent relationship between scheduling schemes, host load, and energy consumption. To tackle this problem, a reinforcement learning-based population hyper-heuristic algorithm is proposed to provide an effective solution. Specifically, six low-level heuristics (LLHs) tailored to the characteristics of the problem are developed and refined. Meanwhile, Q-learning is employed as a high-level strategy for the intelligent selection of LLHs. Moreover, an adaptive local search strategy is introduced to improve inferior individuals. Additionally, an experience-driven task-resource mapping approach is employed to realize efficient resource allocation. To mitigate the challenge of infeasible solutions, a region-constrained insertion mechanism is devised to effectively minimize their occurrence. Subsequently, a solution feasibility reconstruction strategy is introduced to adjust and optimize any remaining infeasible solutions. Extensive experiments on various workflow applications of different scales and types demonstrate the superiority of the proposed algorithm over existing approaches in terms of energy optimization and exhibit significant practical application potential.


  • Open Access

    Interactive Design with Autistic Children using LLM and IoT for Personalized Training: The Good, The Bad and The Challenging

    by Yongfu Wang, Mingyue Tang, Yifan He, Tiffany Y. Tang

    The advent of generative artificial intelligence technologies, such as Large Language Models (LLMs) and Large Vision Models (LVMs), has shown promising results in both academic and industrial sectors, leading to widespread adoption. However, there has been limited focus on applying these technologies to assist children with special needs like Autism Spectrum Disorder (ASD). Meanwhile, conventional personalized training with interactive design for children with special needs continues to face significant challenges with traditional approaches. This workshop aims to provide a platform for researchers, software developers, medical practitioners, and designers to discuss and evaluate the benefits and drawbacks of using LLMs and the Internet of Things (IoT) for the diagnosis and personalized training of autistic children. Through a series of activities, including oral presentations, demonstrations, and panel discussions, this half-day workshop seeks to foster a network of experts dedicated to improving the lives of children with special needs and to inspire further research on leveraging emerging ubiquitous technologies for these underprivileged users, their caregivers and special education teachers.


  • Open Access

    Fitness Landscape Analysis of Genetic Programming Search Spaces with Local Optima Networks

    by Yifan He, Ferrante Neri

    Fitness landscape analysis (FLA) refers to a set of techniques to characterise optimisation problems. This paper presents an FLA of three types of genetic programming (GP) benchmarks: parity, symbolic regression, and artificial ant. We applied a modern graph-based FLA tool called Local Optima Networks and several classical FLA metrics (fitness distance correlation, neutrality, and ruggedness measures) to study the tree-based GP search spaces. Our analysis shows that the search spaces for all problems contain many local optima and are highly deceptive. The parity problems are highly rugged and neutral. Conversely, the problems of symbolic regression are less rugged and neutral. Finally, the artificial ant problem is highly rugged but less neutral. Our results indicate that a mutation in deep nodes makes finding the global optimum difficult.


  • An Energy-Efficient Low-Latency 3D-CNN Accelerator Leveraging Temporal Locality, Full Zero-Skipping, and Hierarchical Load Balance

    by Changchun Zhou, Min Liu, Siyuan Qiu, Yifan He, Hailong Jiao

    Three-dimensional convolutional neural network (3D-CNN) has demonstrated outstanding classification performance in video recognition compared to two-dimensional CNN (2D-CNN), since 3D-CNN not only learns the spatial features of each frame, but also learns the temporal features across all frames. However, 3D-CNN suffers from intensive computation and data movement. To solve these issues, an energy-efficient low-latency 3D-CNN accelerator is proposed. Temporal locality and small differential value dropout are used to increase the sparsity of activation. Furthermore, to fully utilize the sparsity of weight and activation, a full zero-skipping convolutional microarchitecture is proposed. A hierarchical load-balancing scheme is also introduced to improve resource utilization. With the proposed techniques, a 3D-CNN accelerator is designed in a 55-nm low-power CMOS technology, bringing in up to 9.89× speedup compared to the baseline implementation. Benchmarked with C3D, the proposed accelerator achieves an energy efficiency of 4.66 TOPS/W at 100 MHz and 1.08 V supply voltage.


  • Solving Portfolio Optimization Problems Using MOEA/D and Lévy Flight

    by Yifan He, Claus Aranha

    Portfolio optimization is a financial task which requires the allocation of capital on a set of financial assets to achieve a better trade-off between return and risk. To solve this problem, recent studies applied multi-objective evolutionary algorithms (MOEAs) for its natural bi-objective structure. This paper presents a method injecting a distribution-based mutation method named Lévy Flight into a decomposition based MOEA named MOEA/D. The proposed algorithm is compared with three MOEA/D-like algorithms, NSGA-II, and other distribution-based mutation methods on five portfolio optimization benchmarks sized from 31 to 225 in OR library without constraints, assessing with six metrics. Numerical results and statistical test indicate that this method can outperform comparison methods in most cases. We analyze how Lévy Flight contributes to this improvement by promoting global search early in the optimization. We explain this improvement by considering the interaction between mutation method and the property of the problem. We additionally show that our method perform well with a round-lot constraint on Nikkei.


  • Open Access

    Solving Portfolio Optimization Problems Using MOEA/D and Levy Flight

    by Yifan He, Claus Aranha

    Portfolio optimization is a financial task which requires the allocation of capital on a set of financial assets to achieve a better trade-off between return and risk. To solve this problem, recent studies applied multi-objective evolutionary algorithms (MOEAs) for its natural bi-objective structure. This paper presents a method injecting a distribution-based mutation method named L\'evy Flight into a decomposition based MOEA named MOEA/D. The proposed algorithm is compared with three MOEA/D-like algorithms, NSGA-II, and other distribution-based mutation methods on five portfolio optimization benchmarks sized from 31 to 225 in OR library without constraints, assessing with six metrics. Numerical results and statistical test indicate that this method can outperform comparison methods in most cases. We analyze how Levy Flight contributes to this improvement by promoting global search early in the optimization. We explain this improvement by considering the interaction between mutation method and the property of the problem.