Module-3

• Identify the key steps in the AI system development life cycle?

  • Planning
  • Design
  • Development
  • Implementation
• List the key steps in the planning phase?

  1. Consider the business objectives and requirements

  2. Define the business problem

  3. Identify AI use cases

  4. Determine the data needed

  5. Determine the scope

  6. Establish the governance structure
• Describe the key elements in the design stage?

  1. Implement a data strategy, including data gathering and data collection

  2. Examine the quality of the data

  3. Consider data formats (structured vs unstructured, static vs streaming)

  4. Wrangle and prepare the data (cleansing, labeling, anonymization, data minimization)

  
• Data wrangling is the most time-consuming step within the entire development life cycle (about 80% of the entire life cycle) is wrangling/preparing the data
• The five V's of data preparation

  • Volume
  • Velocity
  • Variety
  • Veracity
  • Value
• Describe the steps in data wrangling/preparing the data?

  1. Labeling - tagging or annotating the data to identify what kind of data it is

  2. Anonymization - removing identifiers from the data to protect privacy

  3. Data minimization - not including data that is not needed for the specific application

  4. Privacy-enhancing technologies (PETs) - using techniques like differential privacy and federated learning to protect privacy
• What are PETs and it’s forms like differential privacy?

  Blurs the data by using an algorithm that keeps the data meaningful but makes it nonspecific, so individuals are unidentifiable but the data is still usable
• What is federated learning?

  A way to train models without sharing sensitive data - the global model is in a central location and different locations download it, train it on their local data, and only send the updates back to the central location
• How to determine the system architecture

  1. Choose an algorithm according to desired accuracy and interpretability

  2. Consider requirements and constraints like time, data accuracy, etc.
• Describe the key steps involved in model training, testing and evaluation?

  1. Use representational subsets of the original dataset for training and testing

  2. Train, test, evaluate and retrain different models to determine the best one

  3. Test on relevant evaluation metrics and new data to ensure generalization
• Continuous monitoring


  1. Monitor for deviations in accuracy, irregular decisions, and data drifts that could affect performance

  2. Iterate the model to improve performance as the data changes
• User interviews and market research are two methods used to help identify a business problem that can be solved using AI
• Factors to prioritize AI projects

  • Fit
  • Effort
  • Governance
  • Impact

  
• what is differential privacy?

  Technique that protects information about training data by "blurring" data points using an algorithm to generate values that remain meaningful yet nonspecific
• Once deployed, AI systems require continuous monitoring and maintenance to ensure the model adapts to changes in the environment, especially changes in data
• Data wrangling


  The process of taking raw data and transforming it into a useful format