DRAFT.ipynb

DRAFT¶

Content in this section is based on general ML project pipeline¶

• The tutorial is on the lithofacies predictions from well log data
• It will run through a typical machine learning workflow/pipeline
  • Getting data
  • Data Cleaning
  • Data Visualization and Exploratory Data ANalysis
  • Data Preparation
  • Model Training and Prediction
  • Model Evaluation
  • Model Performance Improvement with different techniques
    • Data Augmentation
    • Cross validation technique
    • Model regularization for better training (for slower weight decay)

Content in this section is focused more on the FORCE competition¶

• It will cover key things to watch out for when working with big subsurface data in the format which we have (the FORCE dataset) e.g.
  • When using gradient boosting trees (possibility of model loss not converging while training)
  • How K-Fold cross validation could help prevent that
  • How selecting a proper cross validation technique helps in making more confident decisions
  • Preventing overfitting by;
    • Creating different validations sets (give "whys" on choice of method for creating them)
    • Making proper evaluations on validation sets with different metrics