Fundamentals of ML and AI

Contents

ML Fundamentals

Training data

  • Labeled data
  • Unlabled data
     
  • Structured data
    • Tabular data
    • Time-series data
  • Unstructured data
    • Text data
    • Image data

Machine learning process

  • Supervised learning
  • Unsupervised learning
  • Reinforcement learning

Inferencing

  • Batch inferencing
  • Real-time inferencing

DL Fundamentals

  • Computer vision
  • Natural language processing (NLP)

General AI Fudamentals

Foundation Models

Foundation Model lifecycle

  • Data selection
  • Pre-training
  • Optimization
  • Evaluation
  • Deployment
  • Feedback and continuous improvement

FMs that are essential to understanding generative AI’s capabilities

  • Large language models (LLM)
  • Diffusion models
    • Forward diffusion
    • Reverse diffusion
  • Multimodal models

Some generative models

  • Generative adversarial networks (GANs)
    • Generator
    • Discriminator
  • Variational autoencoders (VAEs)
    • Encoder
    • Decoder

Optimizing model outputs

  • Prompt engineering
  • Fine-tuning
    • Instruction fine-tuning
    • Reinforcement learning from human feedback (RLHF)
  • Retrieval-augmented generation (RAG)

AWS offering ML Frameworks, AI/ML Services and Generative AI

  • ML Frameworks
    • Amazon Sagemaker - build, train, and deploy ML models for any use case with fully managed infrastructure, tools, and workflows
  • AI/ML Services:
    • Text and documents:
      • Amazon Comprehend - uses ML & Natural Language Processing to uncover insights & relationships of unstructured data, detecting offensive words in text.
      • Amazon Translate - Neural machine translation using deep learning to translate
      • Amazon Textract - extract printed text, handwriting, layout elements, and data from any document
    • Chatbots:
      • Amazon Lex - design, build, test, and deploy conversational interfaces
    • Speech:
      • Amazon Polly - Text to speech, natural-sounding human voices in dozens of languages.
      • Amazon Transcribe - automatic speech recognition (ASR) service for automatically converting speech to text
    • Vision:
      • Amazon Rekognition - image recognition and video analysis with machine learning. can identify objects, people, text, scenes, and activities in images and videos, and even detect inappropriate content.
    • Search:
      • Amazon Kendra - use ML. Intelligent search across organizational data and other webs, etc
    • Recommendations:
      • Amazon Personalize
    • Misc:
      • AWS DeepRacer - to get started with reinforcement learning (RL)
  • Generative AI:
    • Amazon Sagemaker JumpStart - choose one of the language models available and retrain it with your own data for accelerating model development and deployment
    • Amazon Bedrock - Foundation Model as a Service
    • Amazon Q
    • Amazon Q Developer - improve developer productivity

Advantages and benefits of AWS AI solutions

  • Accelerated development and deployment
  • Scalability and cost optimization
  • Flexibility and access to models
  • Integration with AWS tools and services

Cost considerations

  • Responsiveness and availability
  • Redundancy and Regional coverage
  • Performance
  • Token-based pricing
  • Provisioned throughput
  • Custom models