Xiamen University's "Large Model Concepts, Techniques, and Application Practices" (PPT file) - AI Tutorial Materials

《Large Model"Concepts, Technologies, and Application Practices" was produced by the Big Data Teaching Team of Xiamen University and is a popular science book.Large ModelThe technology and its applications in various fields. Emphasis is placed on its influence in the field of big data education. (Review)artificialintelligentA brief history of development, leading toLarge ModelThe definition, characteristics, and classification are explained in detail.Natural Language ProcessingThe document introduces the widespread application of computer vision, speech recognition, and other fields, and its profound impact on work and life.AIGC (artificialintelligentThe content generation technology, including practical applications of text, image, audio, and video generation, as well as assisted programming, was showcased.Large ModelIn content creation,intelligentOfficeAISearch and other aspectspowerfulAbilities. It is a comprehensive, in-depth, and easy-to-understand document.Large ModelPopular science materials, suitable for...artificialintelligentandLarge ModelThis is for readers interested in the technology to study and reference.

Get《Large ModelOriginal PPT file: "Concepts, Technologies, and Application Practices"Scan the QR code to follow and reply: 20250217

Introduction to the Big Data Teaching Team at Xiamen University

The Big Data Teaching Team at Xiamen University is a significant contributor to big data education in Chinese universities. Led by Associate Professor Lin Ziyu, the team members have an average age under 46 and a well-structured composition, encompassing teaching, research, and experimental engineering personnel. Since 2013, the team has focused on big data teaching, demonstrating forward-thinking and...powerfulIts strong execution capabilities have yielded remarkable achievements in areas such as textbook compilation, MOOC courses, and teacher training.

Lin Ziyu is an associate professor in the Department of Computer Science and Technology at Xiamen University. He has authored 15 textbooks in the Big Data series, which have been adopted by over 1,000 universities in China. He has received numerous teaching achievement awards, been selected for several national teaching projects, and been awarded the title of "Typical Application Case of the National Smart Education Public Service Platform of the Ministry of Education." His personal website provides abundant teaching resources and contact information.

Lin Ziyu's "Digital Literacy General Knowledge Course - Big Data and..."artificialintelligent"Computer General Education in the Modern Era" is a textbook for first-year university students, aiming to reconstruct the knowledge system of university computer science courses and cultivate students' computational thinking, data thinking, and...AIThinking skills. The textbook's official website provides a wealth of resources for teachers and students, including lecture slides, MOOC videos, case study videos, computer lab experiments, and teaching outlines.

artificialintelligentBrief History of Development

artificialintelligentIts development can be traced back to the introduction of the Turing Test in 1950. The Dartmouth Conference in 1956 marked...artificialintelligentThe formal establishment of the discipline. After that,artificialintelligentIt has gone through an nascent stage (1950-2005), a period of consolidation (2006-2019), and a period of explosive growth (2020-present). From early CNNs to the Transformer architecture, and then...GPTseries andMultimodalLarge ModelThe appearance ofartificialintelligentTechnological evolution has driven...Natural Language ProcessingComputer vision and other fieldsfastdevelop.

artificialintelligentthinking

• learn:masterAIThe basic principles.
• distinguishClearly define the differences between human and machine capabilities.
• cooperationLearn andAIThe system works collaboratively to improve efficiency.

Large Model:artificialintelligentthe forefront

• Large ModelConcept
  • Large ModelBased onDeep learningThe ultra-large scale of technologyartificialintelligentThe model has a massive number of parameters.powerfulIts learning and generalization abilities are characterized by their large number of parameters, large amount of training data, and high computational resource requirements. For example,GPTThe -3 parameter scale reaches 175 billion, whileGPTThe parameter scale of -4 exceeds 1.8 trillion.
  • Large ModelDevelopment history:Large ModelThe development has gone through three stages:
    • Emerging period (1950-2005): Represented by CNNNeural NetworksModel.
    • Settling period (2006-2019)Models represented by the Transformer architecture.
    • Outbreak period (2020-present):byGPTPre-training seriesLarge Model.
    • Large ModelyesartificialintelligentThis is a cutting-edge technology in the field, belonging to a type of pre-trained model. It is trained on large-scale data and possesses...powerfulLanguage generation, logical reasoning andMultimodalProcessing capacity is the driving forceAIA key force in technological development.
    • Large ModelproductMajor domestic and international projectsLarge ModelProducts include:
      • Foreign products: such as OpenAIChatGPT,Gemini,as well asSorawait.
      • Domestic productsExamples include DeepSeek, Tongyi Qianwen, Doubao, and Wenxin Yiyan.
      • These products areNatural Language Processing,MultimodalIt excels in areas such as understanding and generation, and is widely used in...intelligentScenarios such as customer service, content creation, and code generation.
      • Large Modelprinciple
        • Large ModelBased on the Transformer architecture, and through an encoder-decoder structure and self-attention mechanism, it can process and generate natural language text. Its core capability lies in capturing the complex relationships between words.High efficiencyText generation and understanding.
        • Huge scaleThe number of parameters is huge, and the model size is large.
        • Emergent capabilitiesIt exhibits complex capabilities when trained on large-scale data.
        • Multi-task learningIt can handle multiple tasks and has generalization capabilities.
        • Big Data TrainingIt requires massive amounts of data support.
        • powerfulComputing resourcesIt relies on hardware acceleration such as GPU/TPU.
        • Transfer learning and pre-trainingAdapt to different tasks through pre-training and fine-tuning.
        • Large ModelBased on application areas, it can be divided into three levels: L0, L1, and L2, covering...Natural Language ProcessingIt covers multiple fields such as computer vision and speech recognition.
        • Large ModelApplication areas:Large ModelIt has wide applications in many fields, including but not limited to:
          • Natural Language ProcessingText generation, translation, question-answering systems, etc.
          • Computer VisionApplications include image classification, object detection, and medical image analysis.
          • Speech recognition: Speech synthesis, speech translation, etc.
          • recommendsystemPersonalizationrecommendAdvertising placement, etc.
          • HealthcareApplications include: assisted diagnosis, drug development, etc.
          • Financial risk controlRisk assessment, fraud detection, etc.
          • WorkIt improved work efficiency and promoted...automaticOffice automation andintelligentThe development of customer service.
          • LifeIt has changed the way content is consumed and provided personalized services.
          • Local deployment was introduced.Large ModelIts advantages include data privacy protection, avoidance of usage restrictions, customization flexibility, cost optimization, and offline functionality.High efficiencyUsage. Taking Ollam and DeepSeek R1 as examples, this section demonstrates the specific methods for local deployment.
          • based onLarge ModelofintelligentbodyAble to simulate humansintelligentBehavior possesses autonomy and interactivity. For example, OpenAIOperators and Deep Research are capable of performing complex tasks and providing...High efficiencyPersonalized services.

          AIGC Applications and Practices

          • AIGC Overview
            • AIGC (artificialintelligent(Generated content) is utilizedAIThis technology represents a new way to generate text, images, audio, and video content. Based on techniques such as Generative Adversarial Networks (GANs) and pre-trained models, it can generate high-quality content based on user input and is widely used in creative design, education, entertainment, and other fields.
            • Text classAIGC Application Practice
              • This article introduces the usage methods and techniques of text generation tools (such as DeepSeek and Wenxin Yiyan). These tools can generate articles, poems, code, etc. based on user prompts and support multi-turn dialogue and document processing functions.
              • ImageAIGC Application Practice
                • ImageAIGC utilizationDeep learningImage generation has applications including image creation, restoration, and style transfer. Taking Baidu's Wenxin Yige as an example, it demonstrates how to generate images by inputting...Prompt wordsGenerate high-quality images.
                • VoiceAIGC Application Practice
                  • VoiceAIGC can perform functions such as speech recognition, speech synthesis, and emotion analysis, and is widely used in...intelligentThe field of voice assistants and customer service. Using Doubao and iFlytek as examples, this demonstrates voice-based...AIHow to use GC.
                  • VideoAIGC Application Practice
                    • VideoAIGC can generate video content based on text or images, with applications including advertising production and film special effects. Tencent Smart Film demonstrates how to generate digital human-style narration videos.
                    • AIApplication of GC in assisted programming
                      • AIGC canautomaticGenerate code to improve programming efficiency. Using Doubao (a type of steamed bun) as an example, it demonstrates how to...AIGenerate Python code.
                      • AIsearch
                        • AISearch UtilizationNatural Language ProcessingandMachine LearningTechnology provides more accurate search results. With 360 nanometers...AITaking search as an example, it demonstrates itsMultimodalSearch andintelligentFeatures of tool integration.
                        • AIintelligentOffice
                          • AIintelligentOffice work involves document processing, data analysis, andintelligentFeatures such as assistants improve work efficiency. For example,AIAble to generate initial drafts of documents,automaticFormatting, summarizing document content, and assisting in data analysis.

                          Get《Large ModelOriginal PPT file: "Concepts, Technologies, and Application Practices"Scan the QR code to follow and reply: 20250217

                          How to build a personal knowledge base using DeepSeek-R1 (3 parts)freeTutorial

                          Tsinghua University's DeepSeek andAIHallucination (PDF file) - AITutorial materials