DSSG Hitchhickers guide ML pipeline I

dssg/hitchhikers-guide

DSSG Hitchhickers guide

dssg/hitchhikers-guide

Home
What is in this curriculum?
DSSG Locations
DSSG Manual (for summer fellows)
DSSG Manual (for summer fellows)
Curriculum
Curriculum
- What is in this curriculum
- Sample curriculum for Summer 2019
- Skills you need throughout a project
  Skills you need throughout a project
  - Ethics, Bias, Fairness
  - Communication
    Communication
    
    Intro
    
    Presentations
    
    Writing reports
    
    Visualization
    
    User interface
  - Good Software Practices
    Good Software Practices
    
    Intro
    
    Good repos, good code
    
    Legible, good code
    
    Writing tests
    
    Reproducible software
    
    Pimp my dotfiles!
  - Domain Understanding
- Project Scoping
  Project Scoping
- Project deliverables
- Tech setup
  Tech setup
  - Software you need
  - Software setup session
  - Command line intro
  - Git and github
    Git and github
    
    What is it?
    
    Basic tutorial
    
    Group tutorial
    
    Advanced notes
    
    Git Workflow
    
    Git branching
  - Python
    Python
    
    Basic Python
    
    Python, Pandas and Viz
    
    Python and SQL
  - SQL
    SQL
    
    SQL Basics
    
    Postgres Tips and Pitfalls
  - Good repos
  - Technical workflow
- Getting, storing, and linking data
  Getting, storing, and linking data
  - Intro
  - Data security
  - Get data
    Get data
    
    APIs and scrapping
    
    Working with images
    
    Working with text
    
    Flat files
  - Store data
    Store data
    
    ETL - cleaning, loading
    
    DBs
    DBs
    
    Why a DB?
    
    Designing a DB
    
    Getting data in
    
    Getting data out
    
    Analyzing data (SQL)
    
    Other types of DBs
  - Link data
    Link data
    
    Record linkage
- Data Exploration
  Data Exploration
- Computational and Data Analysis Methods
  Computational and Data Analysis Methods
- Problem Templates in Social Good and Public Policy
- Practical Steps in Using Machine Learning to Solve Social Problems
  Practical Steps in Using Machine Learning to Solve Social Problems
  - Building an ML Pipeline
    Building an ML Pipeline
    
    ML pipeline I
  - Set up problem
    Set up problem
    
    ML problem formulation
    
    ML Checklist
    
    Templates of policy problems
  - Labels/Outcomes
    Labels/Outcomes
    
    One or many
    
    Implications of a label
  - Features/Predictors
    Features/Predictors
    
    Feature engineering
    
    Workshop on feature engineering
    
    Case studies
  - Validation Methodology
    Validation Methodology
    
    Process and goal
    
    K-fold cross-validation
    
    Temporal cross-validation
    
    Field trials
  - Evalution Metrics
    Evalution Metrics
    
    Overview
    
    Examples
  - Models/Methods
    Models/Methods
    
    Machine learning methods
    
    Practical tips on how to use them, parameters, etc.
  - Model selection
    Model selection
    
    Audition
    
    What metrics do we care about?
    What metrics do we care about?
    
    Performance
    
    Stability
    
    Interpretability
    
    Bias
  - Postmodeling
    Postmodeling
    
    Model understanding
    
    Feature importance
    
    Comparing different models
    
    Comparing lists
    
    Error analysis
    
    Bias analysis
  - Experimental design
    Experimental design
    
    Experiment design
    
    Case studies
- Deployment and Maintenance
  Deployment and Maintenance



ML pipeline I

Machine Learning Pipeline
Take a look at Triage, the pipeline we use for most of our projects.
Dirtyduck: A tutorial for triage.