I. Review of Previous Events
Northwestern Polytechnical University started hosting the "Future AI Master" International Summer Camp in 2021. Two editions have been held so far. During the events, internationally renowned scholars from universities such as Yale University, Australian National University, University of Adelaide, City University of Hong Kong, and Northwestern Polytechnical University, as well as prominent companies in the computer industry, were invited to give lectures, forming a top-notch faculty team. More than 100 domestic and international students from 21 universities, including Tsinghua University, University of Science and Technology of China, Beijing Institute of Technology, Harbin Institute of Technology, Tianjin University, University of Electronic Science and Technology of China, and Northwestern Polytechnical University, gathered together. The camps focused on artificial intelligence, internationalization, and leadership, aiming to enhance participants' knowledge and perspectives on AI, overall qualities, and leadership skills through AI-focused learning, project practices, advanced leadership training, and diverse learning and exchange experiences.
II. Camp Background
最近，由人工智能实验室Open AI发布的对话式大型语言模型Chat GPT在全球互联网平台备受关注，它的出现代表着人类在AI领域的不断探索和进步。随着大数据、云计算、互联网、物联网等信息技术的发展，泛在感知数据和图形处理器等计算平台推动，人工智能技术飞速发展，大幅跨越了科学与应用之间的技术鸿沟，在图像分类、语音识别、知识问答、人机对弈、无人驾驶等人工智能技术实现了重大的技术突破，从现有的技术和趋势来看，未来的AI将会更加普及、更加强大、更加智能化。
Recently, Chat GPT, a conversational large-scale language model developed by the AI laboratory Open AI, has garnered global attention on internet platforms. Its emergence represents the continuous exploration and progress of humanity in the field of AI. With the development of information technologies such as big data, cloud computing, the internet, and the Internet of Things, as well as the advancements in ubiquitous sensing data and computing platforms like graphics processors, artificial intelligence technology has been rapidly advancing. It has significantly bridged the technological gap between science and application, achieving major breakthroughs in areas such as image classification, speech recognition, knowledge-based question answering, human-machine games, and autonomous driving. Based on current technologies and trends, AI in the future will become more widespread, powerful, and intelligent.
To provide students with opportunities for collaborative learning and exploration of cutting-edge AI technologies, understanding the new landscape of AI+ era, and further discovering the new value of artificial intelligence, Northwestern Polytechnical University will host the 3rd Future AI Masters International Summer School and the 1st Global IT Competence Excellence Training Camp from July 2nd to July 12th, 2023. （The dates may be adjusted based on the end of exams, tentatively scheduled from July 2nd to 13th）.
III. Camp Features
Feature 1: Invited renowned companies in the computer field such as Huawei, Huawei Cloud, DaoCloud, SOPHGO and Dataa Robotics to offer specialized learning directions. The learning content includes cutting-edge technology foundation courses and high-level competitions. Industry experts from these companies will provide hands-on guidance during practical sessions.
Feature 2: Based on the characteristics of each company, the camp is divided into different learning directions. Participants can choose their preferred company direction to deeply explore the frontiers of artificial intelligence and experience a learning environment that is highly creative, imaginative, innovative, and interactive.
Feature 3: Companies provide computing power support to ensure that participants have practical operating environments and real AI experiences.
Feature 4: The camp provides participants with opportunities to observe and experience AI technology companies on-site.
Feature 5: Some international students can apply for internship recognition.
Feature 6: The camp activities are fun-filled and include lectures, classes, team games, visits to IT industry leaders, and cultural trips to the Tang Dynasty, among other forms of activities.
Feature 7: Outstanding works from Direction 2 can directly participate in the 2023 Shaanxi Regional Competition of the Ascend AI Innovation Contest.
IV: Learning and Practical Project Directions
Direction 1: National Key Projects and Computing for the Nation
· Specific arrangements for this direction will be notified within the QQ group.
Introduction to Some Training Direction:
Dataa Robotics is a unicorn leader in the field of intelligent robotics, a global pioneer, manufacturer, and operator of cloud-based robots. Dataa Robotics has industry-leading full-stack technology solutions for cloud-based robots and has innovatively proposed the "Cloud-Edge-End" ("Cloud AI Brain + Secure Private Network + Edge Deployment + End-side Robot") architecture, successfully commercializing cloud-based robots. In 2022, the Ministry of Science and Technology approved Dataa Robotics to establish the "Cloud-Based Robot National New Generation Artificial Intelligence Open Innovation Platform." HarixRDK, as Dataa Robotics's robot capability development kit, enables the development of robot capabilities in intelligent voice, behavior control, motion editing, and mobile navigation, as well as the synchronization verification of digital twin environments and real robots.
Direction 2: Deep Learning Based on Huawei Ascend Atlas200DK A2
Module 1: Introduction to Artificial Intelligence Fundamentals
课程内容：ChatGPT\文心一言\天工\星火、stable diffusion\dalle2\mid journey等热点应用技术分析、人工智能的定义概念、人工智能的发展历程、人工智能的技术应用分类，人工智能在图像、语言、文本、自动化方向的应用场景和案例，以及人工智能项目的开发流程，包括数据采集标注处理、模型设计训练优化、模型量化剪枝部署等。
Course Content: Analysis of hot application technologies such as ChatGPT, Wenxin Yiyuan, Tiangong, Xinghuo, Stable Diffusion, DALLE2, Mid Journey, etc. Definition and concept of artificial intelligence, the development history of artificial intelligence, classification of artificial intelligence technology applications, application scenarios and cases of artificial intelligence in image, language, text, and automation domains, as well as the development process of AI projects, including data collection, annotation, and processing, model design, training optimization, model quantization, pruning, and deployment.
Learning Objectives: Gain an understanding of the latest technological development trends through practical hot applications, lay a foundation for future learning, stimulate interest in learning, grasp basic knowledge of artificial intelligence, understand the development stage of current artificial intelligence, identify the technical categories of common artificial intelligence applications, recognize the value of artificial intelligence in different application domains, and conduct brief technical analysis of practical cases in various application domains.
Module 2: Artificial Intelligence Neural Network Models
Development of artificial neural networks, full connect neural network (MLP), convolutional neural networks (CNN), recurrent neural networks (RNN), generative adversarial networks (GAN), deep reinforcement learning networks (DRL).
Learning Objectives: Master the role, birth, and evolution of artificial neural networks, understand the structure and operation of artificial neural networks, comprehend the functions of important mathematical functions, be familiar with common neural network models, understand the advantages and application directions of different types of neural network models, and be able to propose suitable network models for solving practical problems.
Module 3: Artificial Intelligence Project Practical Application
(1) Artificial intelligence development environment setup: Anaconda, PyTorch, PyCharm, CUDA, CUDNN.
(2) Drone detection project practical application: Data collection and processing, model construction, model training and evaluation, result testing, deployment on Huawei Atlas200DK A2.
(3) Huawei Cloud developer certification exam: Obtain developer certification awarded by Huawei.
Learning Objectives: Through the target detection model, learn practical application projects of artificial intelligence, master the key points of AI project development, be able to deploy models to Huawei terminal device Atlas200DK, and obtain certification from Huawei.
Direction 3: Introduction and Application of AI Compiler based on MLIR
1.在人工智能领域，为了处理大规模的数据和复的计算任务，硬件加速器成为了不可或缺的工具。Tensor Processing Unit(TPU)作为一种专门针对机器学习任务优化的硬件加速器，具有高性能和能耗效率的优势。为了充分发挥TPU的性能，就需要开发针对TPU优化的编译器。
1. In the field of artificial intelligence, hardware accelerators have become indispensable tools for handling large-scale data and complex computing tasks. Tensor Processing Unit (TPU) is a hardware accelerator specifically optimized for machine learning tasks, offering high performance and energy efficiency advantages. To fully leverage the performance of TPUs, it is necessary to develop compilers optimized for TPUs.
2. TPU-MLIR, as an open-source compiler project, focuses on compiler optimization and tool development for TPUs. It uses MLIR as an intermediate representation language, providing a modular way to define and combine compiler optimizations and transformations. By using TPU-MLIR, a complete toolchain is provided that can convert neural networks from different frameworks into binary files that can be efficiently executed on TPU.
3. To meet the growing demand for compiler development and optimization in the AI field, this course aims to introduce students to the basic principles and techniques of compiler development and apply them in the context of the TPU-MLIR project. Through this course, students will learn how to design, develop, and optimize compilers for achieving the best utilization of TPUs. This will provide students with a deep understanding of compilers and lay a solid foundation for their professional development in the field of artificial intelligence.
1. Understand the basic principles and key technologies of MLIR and AI compilers.
2. Learn about the overall architecture and functionality design of the TPU-MLIR project.
3. Use TPU-MLIR for migrating the ResNet image classification algorithm.
4. Hands-on experience in designing a simple MLIR compiler.
算能致力于成为全球领先的通用算力提供商。算能专注于AI、RISC-V CPU等算力产品的研发和推广应用，以自研产品为核心打造了覆盖“云、边、端”的全场景应用矩阵为城市大脑、智算中心、智慧安防、智慧交通、安全生产、工业质检、智能终端等应用提供算力产品及整体解决方案 。公司在北京、上海、深圳、青岛、厦门等国内 10 多个城市及美国、新加坡等国家设有研发中心。
SOPHGO is committed to becoming the world's leading general computing power provider. SOPHGO focuses on the development and promotion of AI, RISC-V CPU and other computing products. With the self-developed chips as the core, SOPHGO has created a matrix of computing power products, which covers the whole scene of "cloud, edge and terminal" and provides computing power products and overall solutions for urban brains, intelligent computing centers, intelligent security, intelligent transportation, safety production, industrial quality inspection, intelligent terminals and others. SOPHGO has set up R&D centers in more than 10 cities and countries, including Beijing, Shanghai, Shenzhen, Qingdao, Xiamen, the United States and Singapore.
Direction 4: Cloud-Native Edge Computing Platform Practice based on KubeEdge
1. With the rapid development of 5G and the Internet of Things (IoT), more and more applications and data are being deployed and generated at the network edge, highlighting the increasing importance of edge computing. Kubernetes, as the factual standard in the cloud-native container orchestration field, plays a crucial role in edge computing. It can manage and schedule container applications running on edge devices, leading us into a new era of cloud-native edge computing.
2. KubeEdge is a cloud-native edge computing framework designed specifically for edge computing scenarios. It is built on Kubernetes and focuses on providing a consistent experience for cloud-edge resource collaboration, data collaboration, intelligent collaboration, and application collaboration. As the industry's first cloud-native edge computing framework and the only incubating-level edge computing project under CNCF, KubeEdge adopts an open community governance model, connecting the cloud-native and edge computing ecosystems to provide users with integrated cloud-edge-end solutions.
This course aims to introduce students to the basic principles and architecture of KubeEdge, the industry's first cloud-native edge computing platform, and practical case development based on KubeEdge. Through this course, students will learn about the basic principles of cloud-native edge computing, key technologies, and application cases in the industry. This will provide students with the ability to deeply understand edge computing and lay a solid foundation for their career development in the field of cloud-native edge computing.
1. Understand the current status and prospects of the cloud-native edge computing industry.
2. Learn the overall architecture and feature design of the KubeEdge platform.
3. Study typical cloud-edge collaboration production cases of KubeEdge in multiple industries.
4. Develop and deploy a simple cloud-edge collaborative application case based on the KubeEdge platform.
KubeEdge 是业界首个云原生边缘计算框架，云原生计算基金会（CNCF）唯一孵化级边缘计算开源项目。KubeEdge 将Kubernetes 原生的容器编排和调度能力拓展到边缘，为边缘应用部署、云与边缘间的元数据同步、边缘设备管理等提供基础架构支持，实现云边协同、计算下沉、海量边缘设备管理、边缘自治等能力，完整的打通了边缘计算中云、边、设备协同的场景，为用户提供一体化的云边端协同解决方案。
KubeEdge is the industry's first cloud-native edge computing framework and the only incubating-level edge computing open-source project under the Cloud Native Computing Foundation (CNCF). KubeEdge extends the native container orchestration and scheduling capabilities of Kubernetes to the edge, providing infrastructure support for edge application deployment, metadata synchronization between the cloud and edge, and edge device management. It enables cloud-edge collaboration, computing sinking, massive edge device management, edge autonomy, and seamlessly connects cloud, edge, and device collaboration scenarios in edge computing, providing users with an integrated cloud-edge-end collaborative solution.
Huawei, founded in 1987, is a global leader in ICT (Information and Communication Technology) infrastructure and intelligent terminals. Huawei is committed to bringing the digital world to every person, home, and organization, and building an intelligent world of interconnected things.
Shanghai DaoCloud Network Technology Co., Ltd. (DaoCloud) is an innovative leader in the field of cloud-native. Established at the end of 2014, DaoCloud has core technologies with independent intellectual property rights and is dedicated to building an open cloud operating system to empower enterprise digital transformation. Its product capabilities cover the entire lifecycle of cloud-native application development, delivery, and operation, and provide various delivery methods such as public cloud, private cloud, and hybrid cloud.
Direction 5: Deep Learning and TPU Platform Practice in English
深度神经网络模型能很快地被训练、测试，然后被产业界部署从而有效完成现实世界中的任务，在小体积、低功耗的 AI 计算平台上部署这类系统很受产业界欢迎。本课程以实践驱动的方式带领大家直观学习、实战和掌握深度神经网络的知识和技术。
Deep neural network models can be trained, tested, and deployed by the industry quickly to efficiently perform tasks in the real world. Deploying such systems on small-sized, low-power AI computing platforms is highly favored by the industry. This course takes a practical approach to help you intuitively learn, practice, and master the knowledge and techniques of deep neural networks.
The course consists of three main parts. The first part introduces the fundamentals of deep learning and the architectures of CPUs, GPUs, and TPUs. Compared to the widely known general-purpose CPU and GPU computing platforms, TPUs are specialized chips designed for accelerating neural network computations, making them more suitable for deep learning. The second part introduces the TPU chip BM1684, a microsystem-level chip developed by SOPHGO specifically for accelerating neural network operations. The course covers the architecture, performance, and associated toolchains of this chip. The third part is the practical section, covering Chapters Three to Ten, which primarily focuses on model training and the process of running models on the BM1684 platform.
Upon completing this course, students will be able to acquire the following skills:
Fundamental understanding of deep learning
Understanding the differences between CPUs, GPUs, and TPUs
Mastery of the architecture and usage of SOPHGO's TPU chip BM1684, including setting up and using the cross-compilation environment
Learning the implementation of neural networks such as image classification, object detection, semantic segmentation, and instance segmentation on the TPU platform
Basic ability to apply deep learning knowledge to solve specific problems
V. Personal Gains for Students
The summer camp will adopt a content design model of "AI + internationalization + competence + expanded vision" and an implementation model of "modular + project-based + experiential + team-based." Students will have the opportunity to experience high-quality international educational resources, attend cutting-edge subject presentations, and benefit from internationally renowned experts in the field of computer science teaching, fully immersing themselves in and appreciating the charm of artificial intelligence.
All learning and practical activities during the camp will be completed in teams. Each group will be led by two bilingual volunteers proficient in both Chinese and English. Group members come from different countries, regions, schools, and majors. Campers with diverse cultural backgrounds will have the opportunity to learn, live, and grow together.
Students who successfully complete the professional and practical courses will receive highly valued graduation certificates!
Those who perform well have the opportunity to obtain addtional certificates！
VI. Application Process and Timeline
1. Selection Criteria for Campers:
(1) Passion for learning and research, with a strong interest in computer-related technologies such as artificial intelligence, cloud computing, big data, and deep learning.
(2) Proficiency in English communication to a certain extent.
(3) No specific major requirements.
The deadline for camper applications is June 28th.
3 Application Method: Interested students should apply through the on-campus contact persons. The QQ group information can be found below, and you can join the group for inquiries based on your preferred project.
On-campus contact persons: Mrs. Shao, Mrs. Cao
Contact number: 029-88430215
Camp admission notifications (via email and phone) will be sent to qualified applicants by June 30th. Students who do not receive admission notifications by that date have not been selected and will not be notified separately.
The applicant guarantees the authenticity of all application materials.
VII. Registration Fees
For on-campus students: 1000 RMB/person.
For off-campus students: 1800 RMB/person.
（含报名费、学费、活动费、住宿费）(Registration fee, tuition, activity fee, accommodation fee included)