博士研究生岗位 F/M 认知网络可观测性研究：利用图神经网络与大型语言模型实现可泛化且可解释的网络透视技术

该博士项目通过将图神经网络（GNNs）与大型语言模型（LLMs）相结合，来应对网络可观测性挑战。其旨在通过先进的机器学习技术优化网络基础设施管理，重点开展网络透视分析与实时诊断工作。申请者应具备机器学习、网络工程及编程方面的技能。

Assignment: Context and Problem Statement As networks evolve to support ultra-reliable, low-latency communications (URLLC) and complex virtualization paradigms (e.g., 5G/6G Network Slicing), they are becoming increasingly dynamic and opaque. This creates a severe "visibility gap" where operators struggle to diagnose faults without direct administrative access to the underlying infrastructure. Network tomography provides a vital tool to achieve observability by inferring hidden link metrics from end-to-end measurements. However, traditional algebraic and statistical methods suffer from rigidity and scalability issues. Recent advancements have demonstrated that Machine Learning, specifically Relational Graph Convolutional Networks (RGCNs) operating on line graphs, can learn shared link relations and generalize monitor selection. Despite these successes, significant open challenges remain. First, in contrast to most existing literature that relies heavily on synthetic or idealized simulations, there is a critical need to evaluate and generalize these models across entirely different, unseen topologies using real or highly realistic network data. Furthermore, handling non-additive metrics (such as congestion or binary link failures) under these realistic conditions requires more sophisticated architectures. Second, while Graph Neural Networks (GNNs) can accurately infer *where* a degradation occurs mathematically, they lack the administrative and operational context to explain *why* it is happening. This thesis proposes a novel framework that bridges the mathematical inference capabilities of GNNs with the contextual reasoning of Large Language Models (LLMs) to achieve true cognitive network observability. It is conducted within the framework of the GENIE ANR project. Assignments: The PhD student will be responsible for conducting full-time research activities centred on the theme of the thesis: cognitive network observability through the integration of Graph Neural Networks and Large Language Models. The specific assignments include: 1. Advanced GNN Development: Design and implement Relational Graph Convolutional Networks (RGCNs) and novel message-passing paradigms capable of inferring non-additive network metrics and achieving zero-shot transfer to unseen topologies. 2. Graph-RAG Framework Design: Develop a Retrieval-Augmented Generation (RAG) framework tailored for network graphs, including the construction of a vector database containing historical incident tickets, BGP routing logs, maintenance schedules, and vendor documentation. 3. LLM Integration and Evaluation: Integrate the GNN inference pipeline with an LLM equipped with the Graph-RAG architecture, and evaluate the system's ability to generate human-readable, context-aware root cause diagnostics. 4. Experimental Validation: Conduct extensive experiments on real or highly realistic network data to validate the generalizability and robustness of the proposed framework across diverse, unseen topologies. 5. Scientific Dissemination: Write and publish research results in international peer-reviewed conferences and journals, and present findings at scientific events. Main activities: The main activities of the PhD student will include: - **Conducting research** on GNN-based network tomography and LLM-driven explainable diagnostics, including analysis and synthesis of the state of the art. - **Developing experiments** to evaluate model performance across cross-topology generalization and non-additive metric inference scenarios. - **Developing software solutions**, including prototype implementations of the Graph-RAG pipeline and GNN architectures. - **Analyzing research situations** and interpreting experimental results to guide the research direction. - **Writing scientific documents**, including conference papers, journal articles, and the PhD thesis manuscript. - **Presenting research work** at internal seminars, project meetings, and international conferences. - **Collaborating** with project partners within the GENIE ANR consortium and participating in project meetings and collaborative research activities. - **Contributing to the scientific life** of the ERMINE and ADOPNET teams and the Inria Rennes research centre, including participation in seminars and working groups. ## Proposed Research Axes ### Axis 1: Advanced GNNs for Cross-Topology Generalization and Complex Metrics This first axis focuses on pushing the boundaries of GNN-based network tomography to ensure robust performance across diverse, real-world network deployments. - **Zero-Shot Transfer to Unseen Topologies:** Moving beyond generalizing within a single network, this research will investigate zero-shot transfer learning techniques. The objective is to apply models trained on smaller, controlled topologies directly to large-scale, heterogeneous architectures without retraining. - **Non-Additive Metric Inference:** Expanding the GNN's capabilities to infer non-additive metrics, such as bandwidth bottlenecks or binary link states. This will require the development of novel message-passing paradigms capable of capturing complex, non-linear end-to-end relationships. ### Axis 2: Graph-RAG for Explainable Network Diagnostics To transform the raw mathematical inferences of the GNN into actionable, explainable insights, this axis integrates a Retrieval-Augmented Generation (RAG) framework tailored for network graphs. - **Knowledge-Augmented Tomography:** Development of a vector database containing unstructured and semi-structured operational data, including historical incident tickets, BGP routing logs, maintenance schedules, and vendor documentation. - **Explainable Root Cause Analysis:** When the GNN pipeline infers an anomalous metric (e.g., a delay spike) on a specific hidden link, this graph-structured anomaly is fed into the LLM. Using the RAG architecture, the LLM will query the operational database to correlate the mathematical anomaly with real-world events. - **Expected Output:** Instead of a simple inference matrix, the system will generate human-readable diagnostics (e.g., *"The GNN inferred a 40ms delay spike on Link X. Based on retrieved maintenance logs, this correlates with a recent firmware update on the adjacent virtualized router."*). ## Expected Impacts This research will position network observability at the intersection of Network AI and Generative AI. By utilizing GNNs as the "eyes" of the system for inference, and RAG-equipped LLMs as the "brain" for contextual reasoning, this thesis will deliver a highly generalizable, self-explanatory framework. This represents a critical leap toward zero-touch network automation and resilient 6G infrastructure management.

Skills: Machine Learning & Graph Neural Networks: Deep understanding of GNNs, message passing, relational graph convolutional networks, and transfer learning. Network Tomography & Networking: Knowledge of network measurement, performance metrics, 5G/6G, network slicing, and virtualization. Large Language Models & RAG: Experience with LLMs, prompt engineering, vector databases, and Retrieval-Augmented Generation. Programming & Tools: Proficiency in Python, PyTorch/TensorFlow, network simulation tools, and data analysis. Research & Communication: Scientific writing, presentation skills, and ability to collaborate in a multidisciplinary environment. Soft Skills: Autonomy, curiosity, perseverance, and teamwork. The keys to success: Solid ML and networking background: Strong knowledge of Graph Neural Networks, transfer learning, and network tomography, with interest in 5G/6G and virtualization. LLM and RAG experience: Practical skills in Large Language Models, prompt engineering, and building Retrieval-Augmented Generation pipelines with vector databases. Programming and experimentation: Proficiency in Python and deep learning frameworks (PyTorch/TensorFlow), plus rigorous experimental design and analysis. Scientific communication: Ability to write and present research effectively, and to collaborate within a multidisciplinary team. Autonomy and curiosity: Capacity to work independently, explore new directions, and persevere through technical challenges.

福利包： 补贴餐食 部分公共交通费用报销 休假规定：每年7周年假 + 因RTT（法定工作时间减少）额外享有10天假期 + 可能享受特殊假期（如照顾生病的子女、搬家等）。 远程工作可能性（在任职6个月后）及灵活的工作时间安排 提供专业设备（视频会议、计算机设备借用等） 社会、文化和体育活动及项目 职业培训机会 社会保障覆盖范围 薪酬： 税前月薪为2300欧元 申请该职位

Instruction to apply: Please submit online : your resume, cover letter and letters of recommendation eventually Defence Security : This position is likely to be situated in a restricted area (ZRR), as defined in Decree No. 2011-1425 relating to the protection of national scientific and technical potential (PPST).Authorisation to enter an area is granted by the director of the unit, following a favourable Ministerial decision, as defined in the decree of 3 July 2012 relating to the PPST. An unfavourable Ministerial decision in respect of a position situated in a ZRR would result in the cancellation of the appointment. Recruitment Policy : As part of its diversity policy, all Inria positions are accessible to people with disabilities.