Guided Notes
Machine Learning for Informaticists
Supervised vs unsupervised learning, the model lifecycle, features and labels from EHR data, evaluation (discrimination and calibration), bias, drift and what informaticists must ask before deployment.
Types of machine learning
Slide 1 / 5| Type | Learns from | Healthcare example |
|---|---|---|
| Supervised | Labeled outcomes | Predict 30-day readmission; detect pneumonia on X-ray |
| Unsupervised | Patterns without labels | Cluster patient subgroups with similar trajectories |
| Reinforcement | Rewards from actions | Research on treatment strategies (e.g. sepsis fluids) |
| Deep learning | Large data with neural networks | Imaging, waveforms, text |
| Foundation models / LLMs | Massive text (and multimodal) data | Summaries, drafting, extraction |