Special Session Theme: The rapid progress of artificial intelligence, machine learning, large language models, and autonomous agents is creating new opportunities for Electronic Design Automation, especially in complex digital circuit back-end design. As modern integrated circuits continue to scale in complexity, traditional EDA flows face increasing challenges in design-space exploration, placement, routing, timing closure, power optimization, physical verification, and manufacturability-aware optimization. This Special Session aims to explore emerging research at the intersection of Machine Intelligence, Nature-inspired Computing, and Electronic Design Automation, with a particular focus on AI-driven methods for digital circuit physical design and intelligent EDA agents. The session will provide a platform for researchers from academia and industry to discuss how learning-based algorithms, evolutionary computation, reinforcement learning, graph neural networks, large language models, and multi-agent systems can be applied to automate and improve key stages of the EDA workflow. The proposed session will cover both algorithmic advances and practical EDA applications, including intelligent placement and routing, layout optimization, timing and congestion prediction, chip design automation agents, and human-AI collaborative EDA environments.

Scope of the Special Session: The session welcomes original research, case studies, surveys, and practical demonstrations related to AI-driven EDA and intelligent back-end digital design. Topics of interest include, but are not limited to:

Contact Email: 23171214492@stu.xidian.edu.cn

Special Session Theme: Dynamic multiobjective optimization problems (DMOPs) arise in a wide range of real-world systems where objectives, constraints, decision variables, and environmental conditions change over time. Representative examples include intelligent manufacturing, autonomous systems, transportation and logistics, smart grids, communication networks, digital twins, and adaptive resource management. In such scenarios, optimization is no longer a static one-shot task, but a continuous decision-making process that must rapidly adapt to uncertainty, nonstationarity, conflicting objectives, and limited computational budgets.

This Special Session aims to bring together researchers and practitioners working on theories, algorithms, and applications of dynamic multiobjective optimization. The session will provide a focused forum for discussing recent advances in dynamic Pareto optimization, change detection and response strategies, prediction-assisted search, memory-based and transfer-learning mechanisms, surrogate-assisted optimization, evolutionary dynamic optimization, reinforcement learning for adaptive decision-making, and benchmark design for nonstationary environments. Particular attention will be given to methods that improve robustness, responsiveness, scalability, and interpretability under complex and time-varying conditions.

The novelty of this session lies in its emphasis on the intersection of dynamic optimization, intelligent learning, and real-world adaptive systems. Unlike traditional sessions centered on static multiobjective optimization, this session highlights optimization under evolving environments, where both the search landscape and the preferred trade-offs may shift over time. It will encourage contributions that integrate data-driven modeling, online learning, predictive mechanisms, uncertainty quantification, and domain knowledge into dynamic multiobjective frameworks.

Scope of the Special Session: Specific topics for the special session include but are not limited to:

Contact Email: zhenzhongwang@xmu.edu.cn

Special Session Theme: The No Free Lunch (NFL) theorems established a humbling truth for both machine learning and optimization: averaged over all possible problems, no algorithm outperforms any other. The remarkable successes of modern learning and optimization are therefore not the product of universally superior algorithms, but of algorithms whose biases happen to align with the structure of the problems to which they are applied. Effective learning and optimization thus depend fundamentally on incorporating appropriate inductive or search biases, grounded in a deep understanding of the problem domain.

Such biases can be injected at many levels of an algorithmic pipeline. In learning, the hypothesis space is often constrained through specialized architectures, such as convolutional networks for visual data, graph neural networks for relational data, and physics-informed networks that embed conservation laws. Inductive biases also extend beyond architecture to data-centric mechanisms: large-scale pre-training equips foundation models with broad, task-agnostic priors that are subsequently tailored to downstream applications; data augmentation encodes geometric invariances; regularization terms favor desired solution properties; self-supervised objectives shape representation geometry; and Bayesian formulations introduce informative priors. In optimization, search biases manifest in customized operators such as permutation-preserving crossovers and repair mechanisms for constrained spaces, as well as in problem-specific encodings and decodings, landscape-aware initialization strategies, surrogate models that approximate expensive objectives, smoothness-informed predictors for Pareto manifolds, and decomposition frameworks that exploit modularity in large-scale decision spaces. In every case, performance gains stem not from algorithmic universality but from a principled match between bias and problem structure.

This shared philosophy points to a deep and largely underexplored interplay between learning and optimization that this special session aims to highlight. Because both fields succeed by encoding problem-specific knowledge into their inductive or search biases, advances in one can directly inform the other. Learned models can serve as surrogates, heuristics, or operator selectors that inject data-driven biases into search procedures, as seen in neural combinatorial optimization and Bayesian optimization with structured priors. Conversely, search-based methods can shape the inductive biases of learning systems through neural architecture search, evolutionary feature construction, and the discovery of novel loss functions. More broadly, hybrid frameworks that co-design models and search procedures, such as differentiable optimization layers, meta-learned solvers, and the evolutionary construction of learning systems, demonstrate that the boundary between machine learning and strategic search has become remarkably porous.

Scope of the Special Session: This special session invites contributions that engage with bias design as a first-class concern, including but not limited to:

By bringing together researchers from machine learning, evolutionary computation, mathematical programming, and applied domains, the session aims to foster a unified perspective in which the careful design, and, increasingly, the automated discovery, of inductive and search biases is recognized as a central engine of progress in both learning and optimization.

Contact Email: xuexm@upc.edu.cn

Special Session Theme: The rapid advancement of machine learning and large language models (LLMs) is creating unprecedented opportunities for automated algorithm design in complex Black-Box Optimization (BBO). As optimization challenges across engineering, science, and economics grow in scale, traditional heuristic methods face increasing bottlenecks. Constrained by the No Free Lunch theorem, conventional approaches often rely on labor-intensive, expert-driven tuning, leading to limited generalization and sub-optimal performance across diverse problem landscapes.

This Special Session aims to explore emerging research at the intersection of Meta-Learning, Evolutionary Computation, and Black-Box Optimization, with a particular focus on Meta-Black-Box Optimization (MetaBBO). MetaBBO introduces a bi-level, data-driven learning paradigm where a meta-level policy leverages low-level optimization feedback to dynamically guide and configure the algorithmic design of the low-level BBO optimizer. The session will provide a platform for researchers from academia and industry to discuss how advanced learning methodologies can be leveraged to automate and elevate the entire optimization workflow.

The proposed session will bridge foundational algorithmic advances with diverse real-world applications. We particularly welcome submissions demonstrating MetaBBO across domains such as engineering design, intelligent transportation, resource allocation, AutoML, and scientific computing. Ultimately, this session seeks to foster the vision of fully end-to-end autonomous optimization systems that generalize seamlessly across diverse, unseen problem distributions with minimal human expertise.

Scope of the Special Session: The session welcomes original research, case studies, and practical demonstrations related to Meta-Black-Box Optimization and its real-world applications. Topics of interest include, but are not limited to:

Contact Email: gongyuejiao@gmail.com

Special Session Theme: Evolutionary computation (EC) is a powerful class of population-based optimization techniques for solving complex optimization problems. Unlike many conventional optimization methods, EC algorithms generally do not require strong assumptions on problem properties, such as differentiability, convexity, continuity, or explicit mathematical formulations. Owing to their flexibility, robustness, and global search capability, EC algorithms have been widely applied to scientific, engineering, industrial, and data-driven optimization tasks.

Over the past decades, extensive research has been devoted to improving the performance of EC algorithms. Representative efforts include the design of effective evolutionary operators, adaptive parameter control strategies, reliable selection mechanisms, diversity preservation schemes, and hybrid local search techniques. These advances have substantially enhanced the search efficiency and robustness of EC. Nevertheless, when dealing with large-scale, high-dimensional, dynamic, expensive, multimodal, multitask, or highly constrained optimization problems, purely evolutionary search may still suffer from slow convergence, insufficient exploitation of useful search information, and limited scalability.

Recently, learning-aided evolutionary computation (LEC) has attracted increasing attention as a promising paradigm for enhancing evolutionary optimization. The central idea of LEC is to acquire, represent, manage, and reuse useful knowledge extracted from optimization problems, candidate solutions, historical search behaviors, population distributions, environmental changes, or external data sources. By integrating learning mechanisms into the evolutionary process, LEC aims to improve population reproduction, guide search directions, accelerate convergence, maintain population diversity, and reduce computational cost. Various forms of knowledge have been explored in the EC community, including problem-specific knowledge, solution-distribution knowledge, evolutionary-state knowledge, surrogate-assisted knowledge, representation knowledge, transfer knowledge, and knowledge learned from machine learning models, including large language models.

Despite recent progress, LEC remains an emerging research area with many open challenges. First, the theoretical foundations of LEC are still underdeveloped, and rigorous analysis is required to better understand how learned knowledge affects convergence, diversity, population dynamics, and runtime behavior. Second, effective knowledge representation and management remain challenging, especially when the learned knowledge is noisy, partial, dynamic, task-dependent, or expensive to obtain. Third, balancing learning and evolution is a critical issue, as excessive reliance on learned knowledge may lead to premature convergence, while insufficient learning may fail to improve search efficiency. Fourth, there is considerable potential for developing novel LEC frameworks that are scalable, interpretable, adaptive, distributed, and applicable to complex real-world optimization scenarios.

This special session aims to provide a focused forum for researchers and practitioners working on the principles, algorithms, theoretical foundations, and applications of learning-aided evolutionary computation. The session welcomes original research contributions that investigate how knowledge can be effectively learned, represented, transferred, reused, and integrated into evolutionary search. It also encourages studies that develop new learning mechanisms, analyze the behavior of LEC algorithms, and demonstrate the practical impact of LEC on challenging real-world optimization problems.

Scope of the Special Session: Authors are invited to submit high-quality original research papers related to, but not limited to, the following topics:

Contact Email: shwu@scau.edu.cn

Special Session Theme: The rapid convergence of brain-inspired computing, neuromorphic engineering, and embodied intelligence is forging a transformative paradigm for next-generation intelligent systems. While conventional deep learning models and von Neumann computing architectures have achieved remarkable success, they face inherent bottlenecks in energy efficiency, computational latency, and behavioral robustness when deployed in complex, dynamic environments. By mimicking the sparse, asynchronous, and event-driven properties of biological neural circuits, brain-inspired principles offer a promising pathway to overcome these challenges, enabling systems to achieve low-power, high-speed perception, adaptive reasoning, and closed-loop control.

This Special Session aims to provide a premier interdisciplinary forum for researchers from academia and industry to present and discuss the latest advances and emerging trends in brain-inspired intelligent systems, with a particular focus on the interplay between algorithms, neuromorphic hardware, and practical applications. Contributions exploring spiking neural networks, hybrid ANN-SNN computation, neuromorphic learning algorithms, event-based and multimodal sensing, algorithm-hardware co-design, embodied intelligence, robotics, object detection and tracking, and autonomous intelligent systems are particularly encouraged. The proposed session seeks to highlight both theoretical breakthroughs and practical implementations, promoting the development of intelligent systems that are energy-efficient, adaptive, and capable of operating in real-world environments with high temporal fidelity and robustness.

Scope of the Special Session: The session welcomes original research, case studies, surveys, and practical demonstrations related to brain-inspired intelligent systems and their applications. Topics of interest include, but are not limited to:

Contact Email: leideng@mail.tsinghua.edu.cn

Special Session Theme: This special session focuses on brain-inspired computing techniques-including both spiking neural networks (SNNs) and biologically plausible artificial neural networks (ANNs)-for processing and integrating information from multiple sensory modalities (e.g., vision, audio, touch). Brain-inspired approaches offer advantages in energy efficiency, robustness, and adaptive fusion for real-world multimodal tasks. The session aims to bring together researchers working at the intersection of brain-inspired models, multimodal perception, and embodied AI to discuss algorithms, architectures, and applications.

Scope of the Special Session: The session welcomes original research, case studies, and surveys related to brain-inspired methods for multimodal information processing. Topics of interest include, but are not limited to:

Contact Email: maluzhang@uestc.edu.cn

Special Session Theme: This special session focuses on the emerging paradigm of foundation models for brain-computer interfaces, aiming to advance generalizable, robust, and scalable neural decoding and interaction across individuals, tasks, devices, and application scenarios. The session will bring together researchers working on large-scale neural signal modeling, multimodal brain data learning, cross-subject and cross-domain adaptation, and real-world BCI applications.

Scope of the Special Session: This session welcomes large-scale neural signal pretraining, self-supervised and generative neural modeling, unified brain representation learning, prompt-based and parameter-efficient adaptation, zero-shot and few-shot neural decoding, brain-language and brain-multimodal alignment, neural semantic representation, benchmark datasets and evaluation protocols, model scalability and interpretability, and trustworthy deployment of BCI foundation models in healthcare, neurorehabilitation, affective computing, intelligent interaction, and human-AI collaboration.

Contact Email: szhao@zju.edu.cn

Special Session Theme: Neuromorphic computing has emerged as a promising brain-inspired computing paradigm to break through the bottlenecks of traditional von Neumann architecture, including high power consumption, low parallel computing efficiency, and insufficient real-time processing capability for intelligent tasks. Benefiting from its high parallelism, low latency, and ultra-low power consumption characteristics, neuromorphic computing is widely regarded as the core supporting technology for next-generation artificial intelligence, edge computing, and embedded intelligent systems. This special session focuses on the latest research and cutting-edge progress in neuromorphic computing devices and dedicated neuromorphic chips. It aims to gather innovative research achievements covering novel functional devices, brain-inspired circuit design, chip architecture optimization, hardware implementation, and system-level application verification of neuromorphic computing. The session dedicates to providing an academic exchange platform for researchers and engineers in related fields, promoting the cross-integration of microelectronics, material science, artificial intelligence and integrated circuit design, and accelerating the practical deployment and industrialization of high-performance neuromorphic computing hardware systems.

Scope of the Special Session: This session welcomes original research related to the fabrication, physical mechanism, performance optimization, hardware implementation and testing verification of neuromorphic computing devices, sensing-memory-computing integrated devices and neuromorphic chips. The specific topics include, but are not limited to:

Contact Email: xfwang87@uestc.edu.cn

Special Session Theme: Recent advances in brain sensing technologies, artificial intelligence, computational neuroscience, and neuroengineering are expanding the capabilities and applications of Brain-Computer Interfaces (BCIs). Modern BCIs are moving from laboratory prototypes toward intelligent, adaptive, and user-centered systems for communication, rehabilitation, healthcare, assistive control, human-machine interaction, and human augmentation. However, practical BCI systems still face key challenges in signal quality, neural variability, cross-subject generalization, calibration efficiency, real-time decoding, interpretability, usability, and long-term deployment. This Special Session aims to provide a multidisciplinary forum for researchers working on advanced methods, computational models, system designs, and real-world applications of BCIs, covering the full pipeline from neural signal acquisition and processing to intelligent decoding, adaptive interaction, closed-loop feedback, and translational neurotechnology.

Scope of the Special Session: This session welcomes original research, case studies, surveys, and practical demonstrations related to advanced Brain-Computer Interfaces. Topics of interest include, but are not limited to:

Contact Email: caisiqi@hit.edu.cn; songzhenxi@hit.edu.cn

Special Session Theme: Spike cameras provide a novel continuous vision paradigm that captures scene dynamics through ultra-high-speed spike streams. Combined with advances in computational imaging, intelligent perception, neuromorphic computing, and embodied intelligence, spike vision is rapidly expanding from sensing and reconstruction to high-level understanding and real-world applications. Supported by the open-source SpikeCV ecosystem, this Special Session aims to bring together researchers working on spike sensing, imaging, perception, learning, and intelligent systems, and to promote future developments in continuous vision technologies. SpikeCV homepage: https://spikecv.github.io/.

Scope of the Special Session:

Contact Email: yjzheng@pku.edu.cn; spikecv@outlook.com