Labs · Machine learning on well logs

Fit Lab

How machine learning algorithms learn from well logs. In each chapter you try the task by hand, then watch the algorithm do the same thing, one step at a time, on the same samples.

Ten short chaptersSix synthetic wells where the true facies is knownFour real wells to try it on
  1. 0A log is a tableWhat does a model actually see?6 min
  2. 1k-meansCan the logs find the facies on their own?8 min
  3. 2Linear regressionHow well does porosity predict permeability?9 min
  4. 3k-nearest neighboursWhich lithology is this sample, going by the wells we know?8 min
  5. 4Train, validate and testWhy can 98% accuracy be useless?10 min
  6. 5Metrics, plainlyWhich number tells you a model is any good?16 min
  7. 6Decision treesCould a computer choose our cutoffs?8 min
  8. 7Random forestsDo many rough trees beat one careful one?13 min
  9. 8Neural networksCan we predict a missing sonic log?12 min
  10. 9Isolation forestWhich samples are bad hole?10 min

Also in the lab